Experimental Phase Relations in Altered Oceanic Crust: Implications for Carbon Recycling at Subduction Zones

Experimental Phase Relations in Altered Oceanic Crust: Implications for Carbon Recycling at... ABSTRACT The phase relationships in altered mafic oceanic crust (K2O, CO2 and H2O bearing) have been investigated to constrain and quantify the processes of carbon transfer from the slab to the mantle wedge at subduction zones. We report experiments at 2·5–4·5 GPa and 700–900°C in which gas in the experimental charges is analysed by gas chromatography to constrain the volatile composition of the aqueous fluid or hydrous melt. A phengite-bearing epidote-eclogite with dolomite and/or magnesite is stable at subsolidus conditions. The wet solidus at fO2 of NNO (nickel–nickel oxide) was found between 700 and 750°C at 2·5–3·5 GPa and 800–850°C at 4·5 GPa, similar to the CO2-free systems. This observation indicates a low amount of CO2 in the aqueous fluid phase at the wet solidus, in agreement with a measured X(CO2) of 0·059 ± 0·003 at 3 GPa, 700°C and 0·038 ± 0·003 at 3·5 GPa, 700°C. Experiments performed at higher fO2, using either an oxidized starting material or a Re–ReO2 buffer, resulted in a shift of the solidus to higher temperatures. Higher fO2 results in a higher X(CO2) in the aqueous fluid and the reduced water activity leads to a shift in the solidus to higher temperatures. Above the solidus, both textural observations and analyses of the gas enclosed in the capsule suggest that CO2 solubility in the silicate melt increases with increasing P–T conditions. At 4·5 GPa, more than 70% of carbonate at 850°C and 100% at 900°C was dissolved in the hydrous silicate melt. From textural observation it is not clear whether the high carbon content of the melt is related to an increased solubility of carbonate in the hydrous silicate melt or reflects an immiscible carbonatite melt. In any case, partial melting of altered oceanic crust at moderately oxidizing conditions (NNO) and pressures >4 GPa provides an efficient means for the transfer of carbon from the slab to the mantle wedge in intermediate to hot subduction zones. Significant amounts of subducted carbon can thus be brought back to the atmosphere via arc magmatism on relatively short time scales of less than 10 Myr. INTRODUCTION Understanding the deep carbon cycle is important for constraining the evolution of climate over geological time scales (∼100 Myr) (Berner et al., 1983; Edmond & Huh, 2003), the formation of diamonds in the mantle (Harte, 2010; Sverjensky et al., 2014), as well as many processes controlling the chemical evolution of the Earth’s interior (Dasgupta & Hirschmann, 2010; Dasgupta, 2013). Subduction zones play a significant role in the deep carbon cycle as they are the only geological system on Earth that allows carbon to be transported into the deep mantle, which, in turn, is degassed through magmatism (Javoy et al., 1982; Marty & Tolstikhin, 1998). However, whether carbon is effectively recycled towards the atmosphere through arc magmatism located above subduction zones, or buried into the deep mantle, remains controversial, as are the processes involved in carbon mobility within the slab (Fig. 1). The geochemistry of volatiles (mainly the CO2/3He ratio) and basalts (melt inclusions or glass) from arc volcanoes suggests that most of the carbon emitted originates from the subducting slab (Sano & Marty, 1995; Marty & Tolstikhin, 1998; Wallace, 2005). Estimates of the recycling efficiency of carbon at subduction zones are based on the comparison between the carbon fluxes entering subduction zones and outgassing at the surface (Kelemen & Manning, 2015). These estimates differ between studies from 30% to nearly 100% (Varekamp et al., 1992; Marty & Tolstikhin, 1998; Dasgupta & Hirschmann, 2010; Johnston et al., 2011; Kelemen & Manning, 2015) but can be as low as 18–20% (Shaw et al., 2003; de Leeuw et al., 2007; Johnston et al., 2011) (Fig. 1). One source of variability comes from the carbon flux entering the subduction system, which is related to the lithological and chemical composition of the slab, especially the ratio between sediments and altered oceanic crust, the degree of alteration of the oceanic crust and the extent of serpentinization in the subducting lithosphere (Sciurto & Ottonello, 1995; Kerrick & Connolly, 1998; Dasgupta & Hirschmann, 2010; Evans, 2012; Alt et al., 2013). Nevertheless, estimates of carbon fluxes at subduction zones show that the altered oceanic crust is, overall, one of the main hosts of carbon (Staudigel et al., 1989; Alt & Teagle, 1999; Hayes & Waldbauer, 2006; Dasgupta & Hirschmann, 2010; Evans, 2012; Alt et al., 2013; Dasgupta, 2013; Kelemen & Manning, 2015) (Fig. 1). In contrast to geochemical studies, thermodynamic calculations and experimental studies on the stability of carbonate in the subducting altered oceanic crust have mainly yielded contrasting results. Earlier experimental studies conducted on altered oceanic crust suggested that decarbonation mostly occurs at fore-arc depths (Molina & Poli, 2000; Poli et al., 2009) or at much higher temperature conditions (in anhydrous eclogite T > 1050°C and P > 3·5 GPa; Kiseeva et al., 2012). In sediments, decarbonation occurs through the formation of carbonatite at melting conditions of P > 3 GPa, T > 1050°C (Tsuno et al., 2012; Thomsen & Schmidt, 2008), and in carbonate-rich sediments at P = 3 GPa, at T > 1000°C (Skora et al., 2015), conditions that are consistent with only the hottest geotherms of subduction zones or pressures beyond sub-arc depths. At sub-arc depths, experiments (Yaxley & Green, 1994; Molina & Poli, 2000; Hammouda, 2003; Dasgupta et al., 2004; Poli et al., 2009; Skora et al., 2015) and thermodynamic modelling (Kerrick & Connolly, 2001; Gorman et al., 2006; Gonzalez et al., 2016) have shown that carbon is mainly trapped in carbonates or graphite and thus carried into the deep mantle. Fluids released from the subducting altered oceanic crust at sub-arc depths are then expected to have low C/H ratios. Recent work by Poli (2015) has shown that altered plutonic rocks forming the lower part of the oceanic crust experience partial melting at relatively low pressure–temperature conditions corresponding to sub-arc depths (900°C and 3·8–4·2 GPa), with the occurrence of both carbonatitic and silicate melts. This provides an alternative process for C release from the subducted slab. So far, no experiments on altered basaltic composition have been conducted to evaluate whether or not C can be released by melting of altered basalts. Fig. 1. View largeDownload slide Schematic view of the deep carbon cycle in subduction zones. The numbers in black represent the estimation of the yearly fluxes of C (in Tera mols) in and out of subduction zones (Kerrick & Connolly, 2001; Hayes & Waldbauer, 2006; Dasgupta & Hirschmann, 2010; Alt et al., 2013; Dasgupta, 2013; Kelemen & Manning, 2015). Our experimental work shows that along a cold geotherm, 70–90 wt % of C is trapped in residual carbonates. Conversely, in intermediate and hot geotherms, 25–100 wt % stored in the altered oceanic crust can leave the slab via partial melting at sub-arc depth. At low P–T conditions, fluids are produced by dehydration of the subducting lithologies and C is trapped in carbonates (1) (this study). At super-solidus conditions, either a Si-rich melt (2) (this study) or both Si and C melts (3) (this study; Poli, 2015) are produced, allowing for efficient decarbonation of the subducting altered oceanic crust. Schematic representation of textural and solid versus liquid phase relationships at P–T and 1 atm is given on the right-hand side of the figure. Fig. 1. View largeDownload slide Schematic view of the deep carbon cycle in subduction zones. The numbers in black represent the estimation of the yearly fluxes of C (in Tera mols) in and out of subduction zones (Kerrick & Connolly, 2001; Hayes & Waldbauer, 2006; Dasgupta & Hirschmann, 2010; Alt et al., 2013; Dasgupta, 2013; Kelemen & Manning, 2015). Our experimental work shows that along a cold geotherm, 70–90 wt % of C is trapped in residual carbonates. Conversely, in intermediate and hot geotherms, 25–100 wt % stored in the altered oceanic crust can leave the slab via partial melting at sub-arc depth. At low P–T conditions, fluids are produced by dehydration of the subducting lithologies and C is trapped in carbonates (1) (this study). At super-solidus conditions, either a Si-rich melt (2) (this study) or both Si and C melts (3) (this study; Poli, 2015) are produced, allowing for efficient decarbonation of the subducting altered oceanic crust. Schematic representation of textural and solid versus liquid phase relationships at P–T and 1 atm is given on the right-hand side of the figure. When the oceanic crust is altered by seawater, it gains H2O and CO2, but also some K2O (Kelley et al., 2003; Staudigel, 2003), all elements potentially affecting the melting temperature in mafic rocks (Schmidt et al., 2004; Carter et al., 2015). Altered oceanic crust (AOC) is located at the top of the subducting slab, where the temperatures are the hottest (i.e. van Keken et al., 2002) and thus fluid fluxed melting might be possible (Arcay et al., 2005; Syracuse et al., 2010), as investigated in this contribution. Experimental studies are hampered additionally by the difficulty of estimating the volatile composition of the fluid or melt in equilibrium with the solid assemblages composing the metamorphosed altered oceanic crust. Although the volatile composition can theoretically be obtained indirectly from mass-balance calculations, in many cases this method is not capable of resolving small differences in residual carbonates, which inevitably have a large influence on the calculated CO2 content of the fluid phase. To circumvent this problem, analytical techniques allowing the analysis of the gas present in the experimental charge have been developed, including Raman spectrometry (Morgan et al., 1992; Akaishi & Yamaoka, 2000) and mass spectrometry coupled with gas chromatography (Taylor & Foley, 1989; Akaishi et al., 2000; Chepurov et al., 2012; Tiraboschi et al., 2016). In this contribution, we experimentally explore the effect of H2O, CO2 and K2O added during alteration processes on the phase relationships in the AOC and estimate the composition of the gas present in the capsule after the experiment with a gas chromatograph equipped with a thermal conductivity detector (GC-TCD). Our results show the following: (1) at subsolidus conditions, the fluid is water-rich, consistent with previous studies (Kerrick & Connolly, 2001; Poli et al., 2009); (2) the altered oceanic crust partially melts at sub-arc depths conditions; (3) carbonate solubility increases with pressure in the resultant melt, providing a mechanism to transfer carbon from the subducted slab to the mantle wedge. ANALYTICAL AND EXPERIMENTAL METHODS The altered oceanic crust The AOC composition and modes of alteration are mostly inferred from ophiolites (e.g. Gillis & Robinson, 1990) and direct sampling by deep basement drillings (e.g. Staudigel et al., 1981; Alt & Honnorez, 1984; Staudigel, 2003, and references therein). Although deep drilling through young oceanic crust such as hole 504B (Costa Rica Rift, 5·9 Ma) provides information about the early hydrothermal alteration processes (Becker et al., 1989), typical reference composition and style of alteration for the AOC comes from the Ocean Drilling Program (ODP) in older oceanic crust (>100 Ma, such as sites 417/418, 801 and 1149) (Staudigel, 2003). As the mean age for the subducted oceanic crust is older than 60 Ma, these sites provide the AOC end-member component used in discussions about the geochemical processes occurring in subduction zones (e.g. Staudigel, 2003). In the following, we summarize the effect of the alteration on the CO2, H2O and K2O content of the oceanic crust. The oceanic crust experiences important alteration that varies in time and space from the initial cooling at the contact with seawater to its evolution as it moves away from the axial zone (i.e. Staudigel et al., 1981; Fisk et al., 1998; Alt & Teagle, 1999; Staudigel, 2003). In the first 300–600 m, secondary minerals include chlorite, celadonite, saponite, calcite and different Fe-oxides or hydroxides crystallizing in voids or interstices, veins, alteration haloes and fracture fillings (Stakes & O’Neil, 1982; Alt & Honnorez, 1984; Becker et al., 1989; Alt et al., 1996; Alt & Teagle, 1999; Staudigel, 2003). This alteration is the result of complex, subsequent reactions leading to an important uptake of CO2, H2O and K2O (Staudigel & Hart, 1983; Alt & Honnorez, 1984; Staudigel, 2003). In contrast, the deeper parts of the oceanic crust (deeper extrusive and sheeted dykes) experience a more pervasive alteration occurring at higher temperatures, resulting in the crystallization of greenschist- to amphibolite-facies minerals in replacement of primary magmatic phases (Alt & Honnorez, 1984). These secondary minerals, which include pumpellyite, chlorite and calcite amongst others (Staudigel, 2003), reflect enrichment in H2O and CO2 compared with unaltered oceanic crust. Staudigel (2003) noted that the carbon intake of young oceanic crust (sites 504B and 332B) is variable, albeit overall lower than that of mature oceanic crust. Young oceanic crust, such as at site 504B, can thus show different ratios between CO2, K2O and H2O (Alt et al., 1996). An average composition of the altered oceanic crust has been estimated from drill cores, which provide sections of the upper parts of the altered oceanic crust (Alt et al., 1996; Staudigel et al., 1996). The composition of the ‘supercomposite’ was estimated from the mixing of different lithologies in typical proportions found in the altered oceanic crust (Staudigel et al., 1989) and is given in Table 1. Table 1: Composition of the starting materials Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 * Staudigel et al. (1989). In SM3O, all iron is oxidized (see text for details). Table 1: Composition of the starting materials Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 * Staudigel et al. (1989). In SM3O, all iron is oxidized (see text for details). Starting material The compositions of our starting materials are close to the ‘supercomposite’ (Staudigel et al., 1989). However, the K2O content was increased to ∼0·9 wt % to ease phengite identification in experimental runs and to be consistent with more recent estimations of the altered oceanic crust composition (e.g. Kelley et al., 2003). Additional water was added (∼7 wt %) to allow investigation of CO2 release during influx of aqueous fluids from deeper levels of the subducted crust, which is considered the main process for mobilizing slab-derived carbon, major and trace elements (Connolly, 2005; Hermann et al., 2006; Hermann & Spandler, 2008; Plank et al., 2009), and simulate fluid-fluxed melting. Although the altered oceanic crust is likely to experience some decarbonation at lower pressures (<3 GPa; Molina & Poli, 2000; Ague & Nicolescu, 2014), the CO2 content of the supercomposite was kept unchanged to represent an end-member composition for the altered oceanic crust. All starting materials used in the experiments have similar major element compositions. The starting materials SM2 and SM3 were made of synthetic powders mixed in two steps. In the first step, SiO2 as a silica gel, TiO2, MgO together with CaO, K2O and Na2O as carbonates were mixed in an agate mortar. This mix was then pressed into a pellet and fully decarbonated by heating at 800–1000°C in a furnace for 12 h. In a second step, Fe2SiO4, Al(OH)3 and CaCO3 were added to the first mix in the proportions required to obtain the targeted composition. The starting material SM3O was prepared under the same conditions as SM2 and SM3, except that Fe2SiO4 was added before heating at 800–1000°C, ensuring the oxidation of all Fe2+ to Fe3+. The starting material SM2d was obtained by partial dehydration of SM2 in a furnace. The dehydration of Al(OH)3, which is the source of water in the starting materials, was calibrated as a function of time to retrieve the amount of water remaining in SM2d. All three starting materials SM2, SM3 and SM3O were mixed with Li2B2O7 in the proportion 1:4 and then fused to a glass. These glasses were analysed for major elements using a Cameca SX100 operating at 15 kV and 10 nA with a defocused beam at the Research School of Earth Sciences at the ANU. The CO2 and H2O contents of the starting material were calculated from the weight per cent of CaO and Al2O3 measured in the glass and masses of CaCO3 and Al(OH)3 powders added to the starting material. The composition of the starting materials and recalculated CO2 and H2O contents are given in Table 1. Experimental setup Experiments AOC28, 25, 29, 30, 12 and Re4 were designed to estimate the effect of oxygen fugacity on C-bearing phase stability and the phase relations near the solidus (Table 2). Experiments AOC28, 25 and 29 are composed of different proportions (20, 50, 75%) of oxidized (SM3O) and reduced (SM2) starting materials (Table 1), and were run at 3 GPa, 750°C, which is just above the solidus. Experiment AOC30 contained the most oxidized starting material (SM3O) alone. In non-buffered experiments, the assembly (see below) imposes an oxygen fugacity close to NiNiO (NNO; Hermann & Spandler, 2008). The presence of an Fe–Ti oxide, pertaining to the magnetite–ulvöspinel solid-solution in experiment AOC30, suggests that the other experiments made of a mix between reduced and oxidized starting materials have fO2 intermediate between NNO and Magnetite–Hematite (MH) equilibria. Experiments AOC12 and Re4 were buffered by adding a mix of Re–ReO2 at the bottom and top of the capsule. The Re and ReO2 powders were mixed in the proportion 1:5. The Re–ReO2 and SM2 powders were mixed in the proportion 1:1. Experiment Re4 was designed for gas analysis by a gas chromatograph equipped with a thermal conductivity detector (GC-TCD) and thus contained a larger mass of starting material and Re–ReO2 mix (76·7 mg against 36·2 mg in AOC12). The experiments analysed by GC-TCD (AOC40–43, 46 and D850-45) were run with ∼74–95 mg of 0·7SM3 + 0·3SM3O starting material loaded in 3·5 mm inner diameter Au capsules. All the other experiments were run with 2·3 mm inner diameter gold capsules loaded with ∼15–20 mg of SM2 and SM2d starting material. Table 2: Starting materials, P, T and run product in experiments SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – 1 SM, Starting material. 2 30SM3O: Mix of 30%SM3O+70%SM2. 3 Ca-features includes Ca-rich pools and skeletal crystals (DC: dendritic crystals, S: spherules). 4 Fe, Ti: Fe, Ti-rich oxides. 5 Phase proportions calculated with volatile melt composition obtained from GC-TCD analyses (see Table 5) * Phases were identified only. Table 2: Starting materials, P, T and run product in experiments SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – 1 SM, Starting material. 2 30SM3O: Mix of 30%SM3O+70%SM2. 3 Ca-features includes Ca-rich pools and skeletal crystals (DC: dendritic crystals, S: spherules). 4 Fe, Ti: Fe, Ti-rich oxides. 5 Phase proportions calculated with volatile melt composition obtained from GC-TCD analyses (see Table 5) * Phases were identified only. The capsules were then wrapped in a wet tissue to keep them cool and welded shut. The capsule weight was measured before and after the arc welding to detect leaks. For all experiments, the capsule was inserted in a ½ inch assembly made of a MgO-cylinder, a graphite furnace and an external NaCl sleeve. The temperature, accurate to ±5°C, was monitored using a type B thermocouple and a Eurotherm temperature controller. Experiments below 4 GPa were run for 1 week in a 200 t end-loaded piston cylinder at the RSES-ANU. Confining pressure was first applied to the main ram, then the piston was moved at ambient temperature up to ∼0·5 GPa to ensure good compaction of the experimental assembly. Afterwards, pressure was increased to the desired values and temperature was synchronously increased at a rate of 100°C min–1. Experiments at 4·5 GPa were conducted on a 500 t Harwood Engineering press. Pressure and temperature increase were automated and followed the same procedure as before. Piston travel was monitored throughout the duration of the experiments to detect any leak and friction decay (Hermann et al., 2016). All experiments were quenched by shutting off the power to the heater. Mineral identification and major element analyses Phase relations on the mounted and exposed run products were investigated with a scanning Zeiss UltraPlus analytical electron microscope fitted with a field emission gun (FE-SEM) at the Centre for Advanced Microscopy at the ANU. Phase compositions were determined using a JEOL 6400 SEM with energy-dispersive spectrometry (EDS) operating at 15 kV and 1 nA at the Centre for Advanced Microscopy at the ANU. Representative mineral and melt compositions are given in Supplementary DataTable S1 (supplementary data are available for downloading at http://www.petrology.oxfordjournals.org). Quantitative X-ray mapping was performed by wavelength-dispersive spectrometry (WDS) on a JEOL 8530 F electron microprobe at CMCA-UWA with an acceleration voltage of 15 kV and a beam current of 20 nA. The size of the images is 53 × 400 pixels, with a pixel step of 0·3 µm and a dwell time of 100 ms. Mass-balance calculations For each experiment, phase identification and mineral composition analyses were conducted over the whole capsule. Fe- and Mg-rich minerals are typically zoned, but the other elements show reasonably good consistency. For mass-balance calculations, the averaged phase composition was calculated by using analyses performed over the entire capsule. For zoned minerals such as garnet, the average of mineral compositions has been used in the calculation. The mass-balance calculation has been done using the solver function in Excel, by minimizing the difference between the known starting material composition and the recalculated bulk chemical composition from the mineral compositions. GAS ANALYSIS BY GAS CHROMATOGRAPHY Design of GC-TCD for experimental capsule analyses The retrieved gold capsules after the experiments were pierced and the gas was analysed using a GC-TCD. This separates the different components of a gas mix on the basis of their affinity with the stationary phase filling the column and as a function of the parameters chosen for the analysis (Table 3). The instrument is composed of a line and a 0·5 l volume that are connected to a sample holder, a vein-type vacuum pump (VARIAN DS 102), two MKS Baratron® absolute pressure gauges for P > 1 mbar (type 122B) and P < 2 mbar (type 170), a high-purity He cylinder and an Agilent Technologies 6850 gas chromatograph (GC) (Fig. 2a). The sample holder can accommodate either a small gas bottle for standards or experimental charges. The GC is equipped with a capillary column (30 m in length, 0·530 mm internal diameter) filled with HP-PLOT/Q stationary phase and a thermal conductivity detector (TCD). Data acquisition and processing were performed with the Agilent Technologies Chemstation software. Table 3: Parameters of the GC-TCD method for water and CO2 analysis in the experimental samples and standardization Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Table 3: Parameters of the GC-TCD method for water and CO2 analysis in the experimental samples and standardization Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Fig. 2. View largeDownload slide (a) Analytical setup of the gas chromatograph adapted for the analysis of CO2 and H2O in experimental capsules. (b) Chromatogram of the analysis of CO/Air, CH4, CO2 and water balanced with He by GC-TCD with the method parameters given in Table 3. (c) CO2 and H2O characterization of the experimental gas, considered as a proxy for the fluid phase released from the subducting altered oceanic crust. The He-blank is compared with experiment AOC46 (3 GPa, 750°C) and AOC43 (3·5 GPa, 750°C) responses from gas chromatography analysis. Gas has massively exsolved from the melt during the quench of the high-pressure experiment (AOC43) leading to overall larger CO2 and H2O peaks. The X(CO2) of the exsolved fluid is higher in the high-pressure experiment AOC43 compared with experiment AOC46. Fig. 2. View largeDownload slide (a) Analytical setup of the gas chromatograph adapted for the analysis of CO2 and H2O in experimental capsules. (b) Chromatogram of the analysis of CO/Air, CH4, CO2 and water balanced with He by GC-TCD with the method parameters given in Table 3. (c) CO2 and H2O characterization of the experimental gas, considered as a proxy for the fluid phase released from the subducting altered oceanic crust. The He-blank is compared with experiment AOC46 (3 GPa, 750°C) and AOC43 (3·5 GPa, 750°C) responses from gas chromatography analysis. Gas has massively exsolved from the melt during the quench of the high-pressure experiment (AOC43) leading to overall larger CO2 and H2O peaks. The X(CO2) of the exsolved fluid is higher in the high-pressure experiment AOC43 compared with experiment AOC46. Procedure of experimental gas analysis by gas chromatography A typical GC-TCD analysis of an experimental sample lasts ∼5 h. Beforehand, the line and volume were flushed several times with He and evacuated overnight. Vacuum pressures are below 0·0001 mbar. The session starts with the analysis of several blanks (at least four) performed with high-purity He to check for the intensity and reproducibility of the background as well as the presence of possible contaminants. Air and water contamination is systematically observed in the blanks and appears on the chromatograms as a narrow peak eluted at ∼2·2 min for air and a fronting and tailing peak at 6·3 min for water. However, the areas of these peaks are very low (10–25 × 25 µV s) compared with that of the samples (> 250–300 × 25 µV s) and show little variation from one blank to another. The experimental capsule is heated with a heat gun to ∼250°C for 1–2 min to ensure that all free volatiles in the capsule are present as a gas phase. This is especially important for water, which is liquid at room conditions and may remain in the porosity of the experimental charge. Gas from unheated capsules has indeed returned suspiciously high X(CO2). Carbonates from the GC experiments show euhedral shapes and even dendritic crystals were preserved after the heating. Large and small carbonate crystals are equally preserved. Heated experimental charge and unheated experiments performed at the same P–T conditions show the same textures. This demonstrates that no significant carbonate dissolution occurred in consequence of the ∼250°C heating of the capsules. Then the capsule is pierced with a needle under vacuum (Fig. 2). The gas pressure is the difference between the vacuum pressure at room temperature (∼25°C) and the pressure read after piercing and cooling back to room temperature. One and a half hours is needed for the gas pressure in the line to stabilize after heating. The volume and line are topped with He to reach ∼1440 mbar. The gas sample and He mixture is then heated to ensure its homogenization. After cooling to room temperature, the gas mixture is analysed by GC-TCD at least five times. The average area of each peak is calculated with a minimum of five analyses. The relative standard deviation on the average peak area ranges from 0·6 to 2·1% for water and from 0·8 to 1·7% for CO2, attesting to the good reproducibility of the analyses. The validation of the gas analysis by GC-TCD necessitates that all potential gas compounds (CO, CH4, CO2 and water) are eluted at different times and thus are satisfactorily resolved in the chromatogram. For that purpose, a gas mixture composed of 5% CO2, 5% CO, 5% N, 5% O, 4% CH4, 4% H, balanced with He has been analysed with the same method as the samples. The capillary column HP-PLOT/Q used in this study cannot separate air and CO. Repeated He-blank analysis shows that air contamination is very low, ensuring that CO would be identified if present during the analysis. The chromatogram (Fig. 2b) shows that CO, CH4, CO2 and H2O are eluted at 2·222, 2·415, 3·055 and 6·264 min, respectively, which demonstrates that all compounds are satisfactorily separated at the conditions of the analyses. Standardization has been conducted to calibrate the relationship between the peak area and the gas partial pressure in the mixture. Details on the standardization of the gas analyses are given in the Supplementary Material. RESULTS Phase relationships Re-buffered and variable Fe3+ content experiments Experiments with 20–75% of oxidized starting material SM3O were all performed at 750°C, 3 GPa. They show the same phases as observed in experiment AOC13, carried out with SM2 alone (Table 2). In these experiments, a melt coexists with garnet + omphacite + coesite + phengite + epidote + dolomite and rutile is the only Ti-rich phase (Fig. 3a). In contrast, the experiment AOC30, composed of SM3O alone, contains the main paragenesis omphacite + phengite + epidote + kyanite + garnet + coesite + magnesite. In this experiment, melt was not observed. Garnet is present in a low proportion (∼4%, Table 2), magnesite is the sole carbonate present and rutile coexists with an Ti–Fe2+–Fe3+-oxide belonging to the magnetite–ulvöspinel solid solution (Fig. 3b). The experiments AOC12 and Re4, buffered with Re–ReO2, show a mineral assemblage composed of garnet + omphacite + coesite + phengite + epidote + dolomite and rutile (Fig. 3c). Melt was not present in these experiments. Fig. 3. View largeDownload slide BSE images in fO2-controlled experimental run products. (a) Experiment AOC29 with 75% oxidized starting material at 3·0 GPa and 750°C. (b) Experiment AOC30 with oxidized starting material only at 3·0 GPa and 750°C. (c) Experiment AOC12 with Re–ReO2 at 3·0 GPa and 750°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite. Fig. 3. View largeDownload slide BSE images in fO2-controlled experimental run products. (a) Experiment AOC29 with 75% oxidized starting material at 3·0 GPa and 750°C. (b) Experiment AOC30 with oxidized starting material only at 3·0 GPa and 750°C. (c) Experiment AOC12 with Re–ReO2 at 3·0 GPa and 750°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite. Phase relationships of the altered oceanic crust at the experimental conditions Experiments were performed at 2·5 GPa (700–850°C), 3 GPa (700–850°C), 3·5 GPa (700–800°C) and 4·5 GPa (700–1000°C). Run products are all characterized by a typical eclogitic paragenesis composed of garnet, omphacite and rutile (Fig. 4a). The solidus of the altered oceanic crust is located between 700 and 750°C at P ≤ 3·5 GPa and between 800 and 850°C at 4·5 GPa. The formation of a hydrous siliceous melt during the experiments is demonstrated by the presence of a glass sprinkled with many vesicles (Figs 5 and 6). From about 800 to 850°C, the glass is heterogeneous and contains CaCO3 as dendrites, sometimes tens of microns long, fillings of small cavities, or round pools in the melt (Fig. 6). The abundance of these features increases with increasing temperature. At all investigated pressures, the solidus is characterized by a decrease in omphacite, phengite, kyanite (when present), quartz or coesite and epidote, whereas garnet proportion tends to increase (Fig. 7). At P ≤ 3·5 GPa, the solidus coincides with the end of kyanite and magnesite stability fields. Above the solidus, phengite and quartz or coesite are then the next solid phases to disappear. Epidote is present as a major phase only at P = 2·5–3 GPa, 700–750°C. Above the solidus, epidote is mainly an accessory phase and has not been observed at all in the 4·5 GPa experiments. The stability field of dolomite decreases with increasing pressure above the solidus from 700 to 800°C at 2·5–4·5 GPa, where dolomite has been observed only as an accessory mineral in one experiment (AOC47). At P ≤ 3·5 GPa, dolomite is consistently replaced by Mg-calcite above the solidus. At T > 800°C, the solid assemblages are free of water. At P = 4·5 GPa, T > 800°C, the solid assemblage is free of any volatile. Fig. 4. View largeDownload slide (a) Experimental phase relationships in altered oceanic crust. (b) Fluid and melt compositions in the experimental products (see text for explanation). Numbers in black are the mole fraction of CO2 of the quench fluid, obtained from GC-TCD analyses. Numbers in red are the molar per cent of CO3 in the melt estimated from the starting material composition and mineralogy observed in the experiments. Numbers in bold black give the weight percentage of CO2 in the fluid or melt compared with total CO2 in the experiment. Fig. 4. View largeDownload slide (a) Experimental phase relationships in altered oceanic crust. (b) Fluid and melt compositions in the experimental products (see text for explanation). Numbers in black are the mole fraction of CO2 of the quench fluid, obtained from GC-TCD analyses. Numbers in red are the molar per cent of CO3 in the melt estimated from the starting material composition and mineralogy observed in the experiments. Numbers in bold black give the weight percentage of CO2 in the fluid or melt compared with total CO2 in the experiment. Fig. 5. View largeDownload slide Backscattered electron images of experimental mineral assemblages. (a) Subsolidus experiment AOC10 at 2·5 GPa, 700°C. (b) Melt-bearing experiments at 2·5 GPa, 800°C (AOC6). The presence of bubbly glass pools reflecting the exsolution of fluid during the quench should be noted. (c, d) Experiments AOC41 and AOC46 at 3·0 GPa, 700°C and 3·0 GPa, 775°C, respectively. (e) Subsolidus experiment AOC11 at 3·5 GPa, 700°C. (f) Subsolidus experiment D700-45 at 4·5 GPa, 700°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite; Mg-cc, Mg-rich calcite; Lw, lawsonite. Fig. 5. View largeDownload slide Backscattered electron images of experimental mineral assemblages. (a) Subsolidus experiment AOC10 at 2·5 GPa, 700°C. (b) Melt-bearing experiments at 2·5 GPa, 800°C (AOC6). The presence of bubbly glass pools reflecting the exsolution of fluid during the quench should be noted. (c, d) Experiments AOC41 and AOC46 at 3·0 GPa, 700°C and 3·0 GPa, 775°C, respectively. (e) Subsolidus experiment AOC11 at 3·5 GPa, 700°C. (f) Subsolidus experiment D700-45 at 4·5 GPa, 700°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite; Mg-cc, Mg-rich calcite; Lw, lawsonite. Fig. 6. View largeDownload slide BSE images of the Ca-rich features (indicated with the white arrow) observed in association with the glass. (a) Experiment AOC7 at 2·5 GPa, 850°C. (b) Experiment AOC21 at 3·0 GPa, 850°C. (c) Experiment AOC47 at 4·5 GPa, 850°C. (d–f) Mg, K and P X-ray quantitative maps (in element wt %). (g) Experiment AOC38 at 4·5 GPa, 900°C. (h) Experiment AOC34 at 4·5 GPa, 1000°C. In all these experiments, the Si-rich glass is sprinkled with bright, Ca-rich dendritic crystals (DC) or Ca-rich pools (Ca-P), reflecting the high solubility of carbonates in the melt at these conditions. Fig. 6. View largeDownload slide BSE images of the Ca-rich features (indicated with the white arrow) observed in association with the glass. (a) Experiment AOC7 at 2·5 GPa, 850°C. (b) Experiment AOC21 at 3·0 GPa, 850°C. (c) Experiment AOC47 at 4·5 GPa, 850°C. (d–f) Mg, K and P X-ray quantitative maps (in element wt %). (g) Experiment AOC38 at 4·5 GPa, 900°C. (h) Experiment AOC34 at 4·5 GPa, 1000°C. In all these experiments, the Si-rich glass is sprinkled with bright, Ca-rich dendritic crystals (DC) or Ca-rich pools (Ca-P), reflecting the high solubility of carbonates in the melt at these conditions. Fig. 7. View largeDownload slide The systematic change in mineral proportions at (a) 2·5 GPa, (b) 3·0 GPa, (c) 3·5 GPa and (d) 4·5 GPa. The different grey-scale of the markers represents the different starting materials (black, SM2; intermediate grey, 70SM3 + 30SM3O; light grey outline, SM2d). Fig. 7. View largeDownload slide The systematic change in mineral proportions at (a) 2·5 GPa, (b) 3·0 GPa, (c) 3·5 GPa and (d) 4·5 GPa. The different grey-scale of the markers represents the different starting materials (black, SM2; intermediate grey, 70SM3 + 30SM3O; light grey outline, SM2d). Only few experiments were performed at subsolidus conditions. Lawsonite in significant abundance (10%, Table 2), coexisting with magnesite, is restricted to P = 4·5 GPa and T = 700°C. Mineral compositions In all the experiments, most Fe–Mg-bearing phases (omphacite, magnesite, dolomite) and typically garnet show an important zoning in Fe and Mg (Figs 5 and 6d). As a result, the almandine component (XAlmandine) in garnet from one single experiment can vary from 0·65 to 0·40, making it difficult to interpret the Fe–Mg composition of minerals. As analyses on the rims of the grains usually return unreliable results, the XAlmandine in garnet given in this study should therefore be used with caution and considered as a maximum value for the equilibrium composition. Representative mineral compositions are given in the Supplementary Data (Tables S1–S6) and are reported in Fig. 8. Fig. 8. View largeDownload slide Evolution of mineral compositions with P–T conditions in the run products for (a) garnet, (b) omphacite and (c) celadonite. Fig. 8. View largeDownload slide Evolution of mineral compositions with P–T conditions in the run products for (a) garnet, (b) omphacite and (c) celadonite. Garnet Garnet from all experiments except AOC30 does not show significant Fe3+ content. XGrossular shows a general increase with pressure, especially at low temperature: at 700°C, XGrossular varies from 0·27 to 0·39 at 2·5 and 4·5 GPa, respectively. In the SM3O-bearing experiments, garnet has the same grossular content (XGrossular = 0·29–0·31; Fig. 8a) as the SM2-only experiment (AOC13) performed at the same P–T conditions. Garnet in the SM3O-only experiment (AOC30) shows a significantly lower grossular content (XGrossular = 0·19) and contains Fe3+ (XAndradite = 0·10). These garnets are also not zoned in Fe–Mg and do not show any inclusions in their cores. Clinopyroxene In most experiments, clinopyroxene is omphacite with 0·2 ≤ Na/(Na + Ca) ≤ 0·8 and no Fe3+ (Fig. 8) [following the classification of Morimoto (1988)]. AOC30, which contains only the oxidized starting material (100SM3O), is the only experiment showing clinopyroxene with a small component of aegirine (1–7%, 5% on average). The proportion of jadeite in omphacite increases with increasing pressure from 0·22–0·27 at 2·5–3·5 GPa to 0·37 at 4·5 GPa, 800°C. At fixed pressure, the proportion of jadeite in omphacite generally decreases with increasing temperature from 0·27 to 0·21 at 2·5 GPa and 0·37 to 0·33 at 4·5 GPa. In the SM3O-rich and Re-buffered experiments, omphacite crystals show similar jadeite contents (0·26–0·29) to those in the SM2-only experiment AOC13 (0·28). Carbonate phases Mg-calcite is CaCO3-rich (XCa = 0·76–0·82) with minor MgCO3 (XMg = 0·13–0·18) and FeCO3 components (XFe = 0·03–0·05). Dolomite and magnesite are highly zoned in Fe and Mg. Dolomite has an FeCO3 component ranging from XFe 0·15 to 0·31. Magnesite generally has an FeCO3 component ranging from 0·20 to 0·40, and a minor CaCO3 component below 0·06. Dolomite has similar composition in the SM3O-bearing and SM2-only experiments. Phengite The proportion of celadonite in phengite increases with pressure at fixed temperature from 0·4 at 2·5 GPa to 0·6 at 4·5 GPa, all at 700°C. At fixed pressure, the Ti content in phengite tends to increase with increasing temperature, and at fixed temperature, the Ti content in phengite decreases with increasing pressure. Epidote Epidote usually forms needle-shaped crystals that made analyses difficult. In experiments where chemical analyses were obtained, epidote shows a high clinozoisite component (0·67–0·77). At 3·5 GPa, 775°C, the clinozoisite component of the epidote reaches ∼0·92, which suggests that zoisite might actually the stable phase at these conditions. The highest Fe contents in epidote are observed in the Fe3+-rich experiments AOC30 and AOC12, with a clinozoisite component ranging between 0·42 and 0·53, respectively. Fe–Ti-oxides The Fe–Ti oxides observed in experiment AOC30 are composed of Fe2TiO2 (∼0·48) and Fe3O4 (∼0·52), given in molar proportions. Glass and fluid compositions Si-rich glasses Glasses are generally rich in vesicles and the electron microprobe analyses return totals around 80 wt % (Table 4), which attest to their volatile content or porosity. Si-rich glasses have trondhjemite–granodiorite compositions at 2·5 GPa and then become enriched in normative orthoclase with increasing pressure [classification from Barker (1979); Fig. 9]. The SiO2 content of the glasses ranges from 72 to 81 wt % (analyses normalized to 100 wt% without the volatiles H2O and CO2) and Al2O3 from 11 to 17 wt %. The K2O content of the melt ranges from 2·4 to 6·2 wt %. Table 4: Average glass compositions in experimental runs Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 * Melt composition recalculated with CO2 and H2O content measured by GC-TCD. n.d.: not determined. Table 4: Average glass compositions in experimental runs Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 * Melt composition recalculated with CO2 and H2O content measured by GC-TCD. n.d.: not determined. Fig. 9. View largeDownload slide Melt compositions from this study and previous studies in the granite classification of Barker (1979). Fig. 9. View largeDownload slide Melt compositions from this study and previous studies in the granite classification of Barker (1979). Ca-rich pools and dendritic crystals The Ca-rich features are always associated with glass and are very small in size (except at 4·5 GPa, 850°C, few tens of microns; Fig. 6). Their chemical composition, given in Table 4, is usually contaminated by the surrounding Si-melt. However, they show high contents of FeO (3·9–11 wt %) and MgO (1·6–5·6 wt %; Fig. 6d). Their contents of P2O5 (1–2·4 wt %; Fig. 7f), TiO2 (0·3–1·65 wt %) and BaO (detected by EDS analyses) are also higher than what is observed in the silicate melt. They are characterized by very low totals (50–70 wt %), suggesting a high content in volatiles, presumably CO2. Gas In all the experiments analysed by GC-TCD, except Re4, the chromatograms show two peaks at 3·1 s and 6·3 s, which correspond to the elution time of CO2 and H2O, respectively, and a small peak at 2·2 s, which corresponds to air–N2 contamination (area below the peak is <5) (Fig. 2c). The total pressure of gas enclosed in the capsule is typically 7·4–11·6 mbar for 68–94 mg of starting material (Table 5). The mole fraction of CO2 is referred to as X(CO2). GC-TCD analyses of the aqueous fluid present at 700°C show a slight decrease of X(CO2) from 0·059 ± 0·003 to 0·038 ± 0·003 with increasing pressure from 3·0 to 3·5 GPa, respectively (Table 5, Fig. 4b). At 4·5 GPa, 700°C, the capsule leaked during quenching and thus the gas could not be analysed by GC-TCD. Above the solidus at 775°C, the gas present in the capsule is characterized by an X(CO2) of 0·035 ± 0·002 and 0·057 ± 0·003 at 3·0 and 3·5 GPa, respectively (Table 5, Fig. 4b). At 4·5 GPa, 850°C, the gas present in the capsule is characterized by an X(CO2) of 0·043 ± 0·001. In experiment Re4 (3·0 GPa, 750°C), the total pressure of gas in the capsule was 14·4 mbar for 74·3 mg starting material. The chromatogram shows three peaks at 2·2 s, 3·1 s and 6·3 s, which correspond to the elution time of air–N2, CO2 and H2O, respectively. The areas under the CO2 and H2O peaks yield an X(CO2) of ∼0·146 ± 0·006 (Table 5). Table 5: Gas pressure and composition in the experimental runs from GC analyses Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. 1 Pressure of gas in the capsule after piercing under vacuum. 2 Average of area under the peak for each gas. 3 The gas pressure is obtained from the calibration lines established from the standards (see Supplementary Material). 4 XCO2 = PCO2/(PCO2 + PH2O). 5 The mass of CO2 in the capsule is calculated from the ideal gas law. 6 Mass of water in the starting material. n.a., not applicable. Table 5: Gas pressure and composition in the experimental runs from GC analyses Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. 1 Pressure of gas in the capsule after piercing under vacuum. 2 Average of area under the peak for each gas. 3 The gas pressure is obtained from the calibration lines established from the standards (see Supplementary Material). 4 XCO2 = PCO2/(PCO2 + PH2O). 5 The mass of CO2 in the capsule is calculated from the ideal gas law. 6 Mass of water in the starting material. n.a., not applicable. DISCUSSION Approach to equilibrium All experiments contained a fluid phase, facilitating well-crystallized minerals with euhedral shapes. There is no evidence for meta-stable phases and no variation of mineral assemblages throughout the capsule. Phase and melting relations are in agreement with previous studies (see below). At a given pressure, mineral and melt proportions evolve consistently with increasing temperature (Fig. 7), which suggests that the experimental assemblages approached equilibrium. Because of the overall low temperature of the experiments, there is some variation in mineral compositions. Garnet cores are enriched in almandine compared with the rims, suggesting that garnet nucleates early in the experiments, probably eased by the presence of Fe as Fe2SiO4. Nevertheless, garnet rim compositions show a consistent increase in grossular content with increasing pressure from 2·5 GPa to 4·5 GPa (Fig. 8), showing that garnet equilibrated during the experiments. The partition coefficient of Fe and Mg between clinopyroxene and garnet rims overall decreases with increasing temperature at a given pressure and starting material composition. The Si and Ti contents in phengite follow the trend with pressure and temperature described previously by Hermann & Spandler (2008) and Auzanneau et al. (2009). Carbonate stability At subsolidus temperatures, dolomite coexists with magnesite at P = 2·5–3·5 GPa. At 4·5 GPa, magnesite alone is observed with lawsonite at 700°C and then is replaced by dolomite at 800°C (Fig. 10). These parageneses agree with natural assemblages observed in HP and UHP rocks where magnesite and dolomite are typically described in equilibrium with eclogitic assemblages (Compagnoni & Rolfo, 2003; Smit et al., 2008; Zhang et al., 2009). The coexistence of lawsonite and magnesite is also in agreement with thermodynamic modelling on carbonated basalt, where these two phases coexist between 3 and 5 GPa at low temperature (Kerrick & Connolly, 2001). This suggests that our experimental setup and the fO2 conditions imposed by the experimental assembly (close to Ni–NiO equilibrium) yield results that are consistent with natural and modelled observations. By comparison with experiments in the CaCO3–CaMg(CO3)2 system, this study shows that the addition of Fe2+ above 3·5 GPa does not lead to a significant shift of the reaction dolomite + aragonite = Mg-calcite, obtained in the Fe-free system (Hermann et al., 2016). Our experimental results differ from previous work by Poli et al. (2009), in which graphite was found as the main C-bearing phase in the oceanic crust at the same conditions. However, the breakdown of oxalic acid, used in the experiments of Poli et al. (2009) to fix the C/H ratio of the fluid, can involve an excess of H2 or even the crystallization of graphite (Poli et al., 2009; McCubbin et al., 2014), both potentially driving the experiments to lower fO2 conditions. As noted by Poli et al. (2009), carbonate stability and fluid composition are closely linked to oxygen fugacity. Fig. 10. View largeDownload slide Experimental phase relationships for carbonate phases from this study and previous studies. Fig. 10. View largeDownload slide Experimental phase relationships for carbonate phases from this study and previous studies. Controls on oxygen fugacity The experiments AOC 28, 25 and 29 were composed of different proportions of starting materials SM2 and SM3O with different amounts of ferric iron. They all show the same phase relations as the experiment conducted with ferrous iron only (SM2; AOC13) (Table 2). Phase proportions in AOC29 (75%SM3O) are slightly different from those in AOC28 (20%SM3O) and AOC25 (50%SM3O), with lower modes of melt and garnet (Table 2). This suggests that the reducing capacity of the assembly with a graphite heater is very high and is able to counterbalance the addition of Fe3+ up to significant amounts of Fe3+ content (here up to 50–75 wt % of oxidized starting material SM3O). Small amounts of Fe3+ present in the starting material can also be buffered by the major minerals of the HP assemblage such as epidote, omphacite and garnet, without changing the phase relationships. On the other hand, no graphite was found in our experiments, suggesting that oxygen fugacity was not sufficiently low to reduce CO2 to graphite. Also, no Fe loss to the gold capsule was observed, indicating that no significant reduction of FeO occurred. From these observations we conclude that the intrinsic oxygen fugacity in the applied assembly is close to NNO as suggested previously (Hermann & Spandler, 2008). The experiment AOC30, composed of SM3O alone, shows drastically different phase relationships, with magnesite as the only carbonate phase, no melt and Fe–Ti-rich oxides with a high Fe3+ content (Fig. 3b). Garnet is present only as a minor phase (Supplementary Data) and garnet and clinopyroxene display small amounts of andradite and acmite components. Epidote is slightly enriched in Fe compared with experiment AOC13. These observations suggest that the high fO2 imposed by the starting material could no longer be buffered by the experimental assembly or mineral assemblage, but was accommodated by the crystallization of a different mineral paragenesis. Garnet in AOC30 is not zoned in Fe–Mg and does not contain inclusions, which suggests that high fO2 delays garnet nucleation, compared with low oxygen fugacity experiments where abundant fayalite seeds are available. The absence of melt is interpreted to result from the stability of C as CO2 gas in the fluid at high O-fugacity conditions, thus decreasing the activity of water and shifting the solidus to higher temperatures. The higher CO2 content of aqueous fluids at high-fO2 conditions is in line with the estimates of Molina & Poli (2000) comparing fluid composition at Hematite–Magnetite (HM) and NNO buffering conditions in the oceanic crust. O-buffered experiments by the addition of a Re–ReO2 mix show an absence of melt and a low abundance of garnet compared with the other experiments at the same P–T conditions (Fig. 3c;Table 2). In term of gas composition, the pressure of gas enclosed in the capsule is higher (∼14 mbar) compared with non-buffered experiments with higher amounts of starting material (<11·6 mg) (Table 5). The gas is composed of CO2, H2O and an additional gas eluted at 2·2 s. The X(CO2) of the gas is very high (∼0·15) compared with the experiment without Re–ReO2 and performed at 3·0 GPa and 775°C (0·036–0·059) (Table 5). The high gas pressure as well as the presence of an additional gas in the chromatogram compared with Re–ReO2-free experiments suggests that Re combines with OH or C to become volatile at the P–T conditions of the experiments. This shows that Re-compounds are volatile or soluble in water or CO2-rich solutions at high pressures, which is in line with previous observations at P < 0·5 GPa (Chellappa et al., 2009). These experiments confirm that Re can be released in the fluid phase at high pressures in a subduction-zone context (Becker, 2000; Lassiter, 2003). Therefore, the Re–ReO2 buffer is not suitable to buffer fO2 in H2O- or CO2-bearing experiments as it can affect the gas activities and phase relations during the run. The absence of melt in Re4 and AOC12, consistent with previous observations, is interpreted to be the result of high-fO2 conditions produced by the Re–ReO2 buffer and the low water activity. Phase and melting relationships in the altered oceanic crust: comparison with previous experiments and thermodynamic modelling Phase relations in K2O–H2O–CO2-bearing altered oceanic crust The temperature limit of the lawsonite stability field is between 700–800°C at 4·5 GPa and <700°C at 3·5 GPa (Fig. 4a). Although fairly large, this range is similar to that found experimentally or by thermodynamic modelling of basaltic compositions (Poli & Schmidt, 1995; Schmidt & Poli, 1998; Kerrick & Connolly, 2001; Poli et al., 2009). The evolution of modal proportions at 4·5 GPa (Fig. 7d) suggests that garnet grows at the expense of lawsonite, following the de-carbonation reaction lawsonite+4 magnesite+SiO2=pyrope+dolomite+2 CO2+2 H2O. (1) Amphibole has not been observed in this study, which is consistent with the previously determined maximum pressure (∼2·5 GPa) for the stability of this mineral in basaltic systems (Pawley & Holloway, 1993; Poli & Schmidt, 1995). In contrast, epidote is widespread throughout the P–T range investigated in this study, either as a major or minor phase at 2·5–3·5 GPa and T ≤ 750°C or as an accessory phase mostly above the solidus (Fig. 4a). The stability of epidote-group minerals at high-pressure conditions might be enhanced by the presence of trace elements in the starting material. Phengite is the stable K-phase in all subsolidus experiments in agreement with previous studies (Schmidt et al., 2004). Melting and dehydration reactions The presence of melt at experimental conditions is assumed from the presence of a glass in the experimental products. The solidus temperature can be bracketed between 700 and 750°C at 2·5–3·5 GPa (Fig. 4a). The appearance of glass is marked by a decrease of phengite, omphacite and kyanite modes and slight increase in garnet mode with increasing temperature (Fig. 7a–c). Dolomite is the main carbonate phase present below and above the solidus and its mode remains stable throughout the melting process. In contrast, magnesite, present in very minor amounts at subsolidus conditions (<0·03; Table 2), does not survive the onset of melting. It is thus possible that carbonate phases also take part in the melting reaction at 2·5–3·5 GPa. These observations suggest that the melting reaction involves the consumption of phengite, omphacite and kyanite at 3·0–3·5 GPa, and the production of melt and garnet: omphacite+phengite+fluid+coesite±kyanite+carbonate=melt±garnet. (2) The first phases to disappear are kyanite and fluid, as demonstrated by the persistence of coesite, phengite and epidote above the solidus at these pressures. This suggests that there is no excess aqueous fluid phase above the solidus (i.e. the melt is water undersaturated). This is in agreement with textural observations that there are no vesicles with ‘fish egg’ textures as observed in other studies at the same conditions where fluid-saturated melts were present (e.g. Skora et al., 2015). At 4·5 GPa, phengite is strongly enriched in the celadonite component and omphacite in the jadeite component, stabilizing these phases to higher pressure (Kessel et al., 2005) and thus shifting the solidus to higher temperatures. The absence of phengite above the solidus at high pressure is consistent with previous experiments in K2O, H2O-bearing basaltic system showing the coexistence of melt and phengite at 740°C and 3·2 GPa and the absence of phengite at 850°C and 4 GPa (Schmidt et al., 2004). The absence of phengite above the solidus also results in the enrichment of the melt in normative orthoclase, shifting the melt composition into the K-rich granite domain compared with lower pressure experiments from this and previous studies (Fig. 9). Garnet mode above the solidus is relatively constant compared with lower pressures (Fig. 7). Carbonate phases are observed only as accessory minerals at 850°C and disappear completely at higher temperature. This suggests that the melting–dehydration reaction now clearly involves a solid carbonate phase and is omphacite+phengite+dolomite+fluid=melt±garnet. (3) The previous, experimental estimations for the solidus in various chemical systems including either H2O, CO2 or K2O are reported in Fig. 11. The range of solidus temperatures below 3·5 GPa is similar to those observed in K2O, H2O-bearing basaltic compositions (Schmidt et al., 2004; Carter et al., 2015), as well as CO2-bearing, K2O-free basaltic compositions (solidus located between 700 and 750°C between 2·0 and 3·5 GPa; Yaxley & Green, 1994). However, the solidus temperatures found in this study at P ≥ 3·5 GPa are ∼50°C below experimental studies in K2O, H2O-bearing (Schmidt et al., 2004) and K2O-free, H2O-bearing basaltic compositions (Kessel et al., 2005). This might be explained by the lower water content in the first study (1·8 wt % compared with 6–7 wt % in this study) and the absence of K2O and phengite in the second study (Kessel et al., 2005). The solidus temperature determined here at 3 GPa occurs at a slightly lower temperature (700–750°C) than in the experiments of Skora et al. (2015) on low-carbonate pelite (LC) containing ∼7 wt % water (760–800°C) and those performed by Carter et al. (2015) (∼750°C, 3 GPa) on an altered oceanic crust composition (K2O = 0·14 wt %). For Skora et al. (2015), this difference in the solidus temperature is well explained by the higher Fe3+ content of their starting material (Fe3+/Fetot = 0·46), combined with a different experimental assembly buffering at slightly higher fO2 (above NNO). The experiments performed at different fO2 shown in this study indeed demonstrate that fO2 is critical in controlling the onset of the melting reaction, and experiments with the highly oxidized starting material (AOC30) or the experiment buffered at high fO2 (AOC12) show a solidus shifted to higher temperature conditions. The slight difference in melting temperature compared with the experiments of Carter et al. (2015) can again be assigned to differences in fO2 conditions, as demonstrated by the occurrence of Fe–Ti oxides and the difficulty of nucleating garnet in their experiments. Our results differ from the previous experiments on H2O–CO2–basaltic systems by Poli et al. (2009), where only one experiment, performed at fO2 conditions of NNO, showed a glass at 2·2 GPa, 730°C. The low solidus temperature found in our study, which is consistent with CO2-free and CO2-bearing systems at least at P < 4 GPa (Yaxley & Green, 1994; Poli & Schmidt, 1995; Schmidt et al., 2004), shows that the presence of CO2, when trapped in carbonate phases below the solidus, does not affect significantly the water activity in the system. This contrasts with high-fO2 conditions, which lead to dissolution of carbonates and fluids with higher amounts of CO2, and thus an increase in the solidus temperature. Fig. 11. View largeDownload slide Comparison of solidus temperatures for mid-ocean ridge basalt, altered oceanic crust and pelite systems. For intermediate to hot subduction geotherms (Syracuse et al., 2010) the subducted altered oceanic crust is able to undergo fluid fluxed partial melting at sub-arc depths (3–4·5 GPa). P–T paths of the slab surface beneath the Lesser Antilles (a) and Sunda (b) arc segments are from Syracuse et al. (2010). Fig. 11. View largeDownload slide Comparison of solidus temperatures for mid-ocean ridge basalt, altered oceanic crust and pelite systems. For intermediate to hot subduction geotherms (Syracuse et al., 2010) the subducted altered oceanic crust is able to undergo fluid fluxed partial melting at sub-arc depths (3–4·5 GPa). P–T paths of the slab surface beneath the Lesser Antilles (a) and Sunda (b) arc segments are from Syracuse et al. (2010). Nature and H2O–CO2 composition of the fluid phase at experimental conditions Aqueous fluid At 700°C (3–4·5 GPa), the volatile-rich phase is close to an aqueous fluid (scheme 1, Fig. 1), as demonstrated by the high water content of the gas escaping the capsule [X(CO2) = 0·038–0·059 at 3–3·5 GPa; Table 5] and the absence of glass in the experimental charges. The amount of fluid released determined by GC-TCD analyses indicates that about 70 wt % (at 3 GPa) and 80 wt % (at 3·5 GPa) of the carbon present in the starting material was retained in the residual assemblage (Fig. 4b). At 4·5 GPa, 700°C, the capsule leaked during quenching and thus the gas could not be analysed by GC-TCD. Nevertheless, SEM investigation of the charge, as well as mass-balance calculations, shows that a high amount of magnesite was present (∼5%; Table 2) and thus ∼90 wt % of the C was retained in the residue (Fig. 4b). The calculated X(CO2) is very low at 0·02–0·03, following the trend established at lower pressure. These results are in excellent agreement with calculated fluid compositions based on experiments that predict a decrease of X(CO2) in the fluid below ∼0·05 with increasing pressure above 2·5 GPa at NNO-buffered oxygen fugacity (Poli et al., 2009). Thermodynamic modelling of fluid compositions provided significantly lower X(CO2) ≤ 0·01 at sub-arc depths in an altered basalt composition (Connolly, 2005). The thermodynamic modelling was performed with a mixing model between molecular CO2 and H2O, which was consistent with experimental studies below 1·6 GPa on calcite solubility in water, showing that CO2aq was the main carbon-rich species (Caciagli & Manning, 2003). However, the higher solubility of carbonate in our experiments might suggest the presence of dissolved carbonate ions in the complex aqueous fluid at the higher pressures of our run conditions. This is supported by evidence of carbonate dissolution from fluid inclusion studies in natural UHP rocks (Frezzotti et al., 2011) and by recent thermodynamic modelling of C speciation in high-pressure fluids in equilibrium with eclogitic minerals (Sverjensky et al., 2014). To our knowledge, our data represent the first direct measurements of the CO2 concentration in aqueous subduction-zone fluids. Hydrous silicate melt The observed phase relations indicate that a single hydrous silicate melt (and no excess aqueous fluid) was present above the solidus. The volatile content of this melt can be constrained from the amount of observed H2O- and CO2-bearing phases (Table 2) and from the quantitative analysis of gas. The mass of volatiles coming out of the capsules, obtained from the GC-TCD analyses, can be used to estimate the proportion of gas exsolved during the quench. A minimum estimate is calculated assuming that there is no more CO2 in the glass. A maximum estimate considers that all the carbonates were dissolved in the melt. These calculations show that 60–80 wt % of gas was exsolved from the melt at 3·5 GPa and 40–58 wt % at 3·0 GPa. This observation demonstrates the importance of analysing the fluid composition by direct methods, here by GC-TCD, to well constrain the volatile budget of the experiments, as the glass composition may not represent the composition of the melt. At T > 700°C, the exsolved gas is characterized by an X(CO2) of 0·035 ± 0·002 and 0·057 ± 0·003 at 3·0 and 3·5 GPa, respectively (Table 5, Fig. 4b). It is more difficult to assess the CO2–H2O composition of the hydrous siliceous melt than of the fluids at subsolidus conditions because it is very difficult to measure the volatile composition of the small, bubble-rich glass pools by any analytical technique. The minimum X(CO2) in the hydrous siliceous melt can be estimated from the GC-TCD measurement and considering CO2Glass = 0, as X(CO2)Melt=CO2Gas/44H2OGlass/18+H2OGas/18+CO2Gas/44. H2OGas and CO2Gas are the mass of water and CO2 in the gas phase and are calculated from GC-TCD analysis by converting the gas partial pressure to mass using the ideal gas law. H2OGlass is obtained by mass balance using the proportion of hydrous siliceous melt and hydrous minerals in the experiments and the total mass of water in the starting material. The resulting minimum X(CO2) of the hydrous siliceous melt present at experimental conditions is thus >0·020 at 3·0 GPa and >0·045 at 3·5 GPa, 775°C (Fig. 4b). These values are lower than the bulk-rock X(CO2) of these experiments (0·16), which is consistent with the observation that residual carbonates are present in these runs (Table 2). These X(CO2) values translate to a proportion of CO2 released in the fluid phase (mass of CO2 in the fluid or melt divided by the mass of CO2 in the starting material) ranging from >10 to >25% (Fig. 4). Increased solubility of CO2 in silicate melts with pressure and implication for fO2 in the mantle wedge Some of our super-solidus experiments show Ca- and CO2-rich features associated with the silicate glasses. Their origin can be (1) a quench product from an immiscible carbonatite liquid (scheme 3, Fig. 1), (2) an unmixing of dissolved carbonate liquid on quenching of a single carbonate-rich silicate-melt (scheme 2, Fig. 1), or (3) quench material of solutes that derive from the unmixing of an aqueous fluid phase from a melt during quenching. The description of different quench textures was briefly reviewed by Poli (2015). Previous experimental work suggested that dendritic crystals tend to form as a quench product of a liquid (Wyllie & Tuttle, 1959). Quench carbonatite melts usually form large spherical pools with their own internal organization of dendritic crystals (Martin et al., 2013; Skora et al., 2015). In contrast, quench material after an exsolved fluid appears in spherical micro-vesicles sprinkled throughout the glass (Skora et al., 2015). At 850°C, 4·5 GPa, only accessory (<2%) residual carbonates and phengite (only in D850-45) were observed. The glass contains long Ca-rich dendritic crystals (Fig. 6) characterized by relatively high P, Ba and Ti contents. These dendritic crystals are interpreted to have formed from a carbonatite liquid exsolved from the hydrous silicate melt during quenching. In the higher temperature experiments at 4·5 GPa, dendritic crystals are replaced by micro-vesicles filled with Ca-rich material, suggesting that a gas, rather than a liquid, might have been exsolved during the quench. The textures observed in this study are very similar to those reported by Poli (2015) in experiments on hydrous carbonated gabbro at 4·2 GPa, 870–900°C. It remains unclear whether a single, hydrous silicate melt with a significant carbonatite component or two separate melts existed at the experimental conditions. In any case, all these observations reflect an enhancement of carbonate solubility in the hydrous siliceous melt. At 4·5 GPa, 850°C, the near total dissolution of all the carbonate and the large proportion of the dendritic crystals in the glass pools (10–15%) indicates that 75–100 wt % of C from the starting material was hosted in the hydrous siliceous melt, resulting in an estimated X(CO2) > 0·08. At 4·5 GPa, 900°C, all the carbonate added to the starting material is dissolved in the hydrous siliceous melt coexisting with garnet + omphacite + coesite. This experiment was performed with a lower water content [X(CO2) = 0·27] to avoid formation of an excess fluid phase. As no carbonate mineral is present in this experiment, the X(CO2) of the melt is 0·27 and 100% of the total CO2 present in the starting material occurs in the hydrous siliceous melt (Fig. 1). It is worth noting that the increase in X(CO2) of the hydrous siliceous melt from >0·08 to 0·27 at 4·5 GPa, 850–900°C reflects the composition of the starting material and, thus, the increased capacity of the hydrous siliceous melt to incorporate carbonate species. The increase of carbonate solubility with pressure and temperature in hydrous siliceous melts is thus dramatic at P > 4 GPa. The carbonate-rich melts produced in the slab at sub-arc pressures are also likely to invade the mantle wedge. The oxidizing power of CO2 in melts was demonstrated from experiments (Simakin et al., 2012) and studies of natural magmatic systems that evolved in response to interaction with carbonate-rich rocks (Wenzel et al., 2002; Simakin et al., 2012). In these studies, the increase in fO2 of the melt results in Fe oxidation and crystallization of Fe3+-rich clinopyroxene or Mg-rich olivine. This is consistent with the high Fe content of the C-rich features observed in our study, compared with the Si-rich glasses (Table 4). CO2-rich melts are thus potentially an efficient transfer medium for oxidized species and could increase the input of Fe and C in subduction zones. The oxidation state of the sub-arc mantle has long been a subject of debate (Parkinson & Arculus, 1999; Aeolus Lee et al., 2005) and has strong implications in terms of melting relations (Foley, 2011), redox budget (Evans, 2012; Evans et al., 2012) and the generation of Cu–Au deposits (e.g. Evans & Tomkins, 2011). Evidence for an oxidized sub-arc mantle includes primitive arc magma compositions and fO2 estimations from minerals forming arc basalts and associated peridotites (e.g. Parkinson & Arculus, 1999; Rowe et al., 2009; Evans et al., 2012). The oxidation of the sub-arc mantle is interpreted as the result of interaction with oxidized fluids coming from the subducting slab and usually considered as aqueous or silica-rich fluids (Evans, 2012). Our study suggests that CO2-rich melts formed at sub-arc depths in the subducting altered oceanic crust could also be responsible for the oxidation of the sub-arc mantle. IMPLICATIONS FOR CO2 AND H2O TRANSFER AT SUBDUCTION ZONES Our study suggests that the three critical factors controlling carbon recycling at subduction zones from altered oceanic crust are oxygen fugacity conditions, flushing of aqueous fluids from the dehydration of underlying units and the temperature on top of the slab at sub-arc pressures. The alteration of the oceanic crust evolves with depth from oxidizing to rather reducing conditions (Alt & Honnorez, 1984; Staudigel, 2003). Nevertheless, the mineral parageneses and compositions in blueschist and eclogite rocks generally suggest that high-pressure metamorphism, and associated devolatilization, then drives the fO2 towards reducing conditions. Rutile is the most common Ti-rich phase in high-pressure metabasites, whereas ilmenite or other Fe–Ti oxides are rarely observed unless at very low degrees of metamorphism (Galvez et al., 2012). In addition, garnet rarely shows a high andradite component. The study of Groppo & Castelli (2010) interpreted the Fe3+/ΣFe zoning in garnet as a decrease in fO2 across the FMQ buffer during prograde lawsonite destabilization in metabasites. Cao et al. (2011) estimated the fO2 of massive eclogite from the North Qilian suture zone (NW China) at ΔFMQ +2·5. The good consistency between the parageneses observed in our experiments and in natural rocks, as well as fO2 estimations in natural eclogites, suggests that the fO2 conditions of our experiments are relevant to subducted, altered oceanic crust at sub-arc conditions. Nevertheless, some occurrences of very oxidized rocks, associated with a peculiar mineralogy, are locally described in HP complexes; for example, the Mn ore of Praborna in the Italian Western Alps formed at 2 GPa, which records ΔFMQ up to 12·7 (Tumiati et al., 2015), and the epidote-talc bearing eclogites from the Sulu Terrane, which record fO2 conditions at ΔFMQ = 4·5 (Mattison et al., 2004). In such rare rock types, CO2 extraction owing to fluid–rock interaction is expected to be much more efficient. In the past, fluid-present melting of altered basalts has not been regarded as feasible because thermal models predicted temperatures too low for this to happen at sub-arc depths (Peacock, 1996). However, in the newest set of thermal models (Syracuse et al., 2010), a large number of subduction zones reach temperatures where the top of the altered basalt layer might undergo fluid-fluxed partial melting (Fig. 11). In addition, the wet solidus of altered basalt determined here is very similar to the wet solidus of sediments. Studies on K, Th and light rare earth element recycling in subducted sediments have demonstrated that partial melting is necessary in most subduction zones, suggesting that top slab temperatures can reach 700–900°C (Hermann & Spandler, 2008; Plank et al., 2009) and supporting partial melting in the altered oceanic crust as a likely process in subduction zones. Our experiments constrain water-fluxed melting of one specific type of altered oceanic crust, a K-enriched ‘supercomposite’ (Staudigel et al., 1989). The amount of carbonate and K-phases is highly variable depending on the depth and processes of metasomatism, as summarized in the section above on alteration of the oceanic crust. Although these highly variable rock compositions have not been investigated in this study, the obtained phase and melting relations shed some light on the devolatilization of these rocks. The solid phases at subsolidus conditions just before the onset of melting are garnet, clinopyroxene, coesite, phengite, carbonate ± kyanite, in agreement with a wide range of composition between sediments and basalts (Schmidt et al., 2004; Hermann & Spandler, 2008; Carter et al., 2015). Thus, different bulk compositions will result only in different proportions of these minerals (with variable compositions) but will not modify melting reactions (2) and (3). The extent of melting will then depend first on the amount of external aqueous fluid input and second on the order of exhaustion of the solid phases participating in the melting reactions. The melting temperature is expected to vary with phase composition, particularly the Mg# of the bulk-rock and the Na/Ca of the omphacite. In the absence of any external fluid input at sub-arc depths, the extent of melting of altered oceanic crust will be controlled by the amount of phengite and epidote, as they are the only stable hydrous phase at the solidus of these rocks. In this scenario the efficiency of carbon removal from oceanic crust will depend on the ratio between phengite + epidote and carbonate, and thus on the extent and type of alteration of the oceanic crust. The most recent estimations of carbon fluxes in subduction zones by Kelemen & Manning (2015) suggest that most carbon subducted along with serpentinite, sediments and the altered oceanic crust is likely to be brought back into the mantle wedge via fluids migrating from the slab. The main process proposed by the above researchers is the dissolution of carbonate minerals, calcite–aragonite, into the aqueous fluids percolating through the different lithologies of the slab. Our experimental results, combined with other experimental studies at similar conditions (Carter et al., 2015; Poli, 2015; Skora et al., 2015), demonstrate that partial melting of the subducting, crustal lithologies is possible at least along warm and hot geotherms (Fig. 11) and constitutes an alternative, effective way to mobilize carbon from the slab towards the mantle wedge. Furthermore, the model put forward by Kelemen & Manning (2015) necessitates diffusive transport of the percolating aqueous fluids throughout the slab lithologies, which is difficult to reconcile with natural observations showing that aqueous fluid flow is rather channelized than pervasive (e.g. Spandler et al., 2011). In terms of fluid transport, carbonatite melts are characterized by very low viscosity, especially at high-pressure conditions (e.g. Kono et al., 2014). The dissolution of carbonate in hydrous silicate melts would probably decrease their viscosity, thus helping in the migration of those fluids from the top of the slab to the mantle wedge. The time scales of deep carbon cycling critically depend on the main process of carbon release. Volatile recycling through arc magmatism takes less than 10 Myr (Morris et al., 1990). Residual carbonate in eclogite may result in melting at the Transition Zone or even at lower mantle conditions (Rohrbach & Schmidt, 2011) and recycling through intra-plate magmatism takes of the order of 100 Myr. The time scale for subduction to the lower mantle and recycling through plumes and mid-ocean ridge magmatism probably exceeds 1 Gyr (Dasgupta & Hirschmann, 2010). As altered oceanic crust is the main reservoir of subducted carbon, the efficient CO2 extraction by partial melting of subducted altered oceanic crust over short time scales of <10 Myr has important consequences for the long-term evolution of CO2 in the atmosphere (Berner et al., 1983). Interestingly, the range of X(CO2) of ∼0·02–0·08 mol % produced during wet melting at sub-arc depths (Fig. 4b) translates into a range of H/C mass ratios of ∼1–3. These values straddle the average H/C mass ratio of 1·92 of the exosphere (Hirschmann & Dasgupta, 2009). It needs to be evaluated, with future studies, whether wet melting of carbonated eclogite has been a dominant process throughout Earth’s history that could have fractionated C and H in the way that they are now present in the exosphere or whether this is just a coincidence. CONCLUSIONS (1) Solidus temperatures for the altered oceanic crust are between 700 and 750°C at 2·5–3·5 GPa and 800–850°C at 4·5 GPa, similar to those for CO2-free, K-bearing basaltic systems. This provides evidence for low solubility of CO2 in aqueous fluids in equilibrium with carbonate-bearing eclogite residues. (2) We have directly measured such aqueous fluids using gas chromatography, confirming a low X(CO2) of 0·059 ± 0·003 at 3 GPa, 700°C and 0·038 ± 0·003 at 3·5 GPa, 700°C. Although these values are low in absolute terms, they are still significantly higher than calculated values at the same conditions, which are generally below 0·01. This indicates that fluids are probably complex and might have also significant amounts of dissolved carbonate ions at sub-arc conditions. (3) Different proportions of oxidized versus reduced materials, as well as Re–ReO2, have been used to investigate the effect of fO2 on carbonate stability in the altered oceanic crust. It is shown that the reducing power of the assembly set the fO2 around NNO, even when up to 75% of oxidized material is used in the starting material. Experiments buffered with Re–Re2O suggest that Re gas compounds mix with CO2 and H2O, confirming that Re can become volatile at subduction conditions. An experiment carried out with oxidized material suggests that high fO2 stabilizes CO2 gas over carbonates at subsolidus conditions, thus increasing the solidus temperature at sub-arc depths and delaying the onset of partial melting to higher temperatures in subduction zones. (4) Partial melting of altered oceanic crust is possible in subduction zones along intermediate to hot geotherms for fO2 close to NNO. Carbonate solubility in these hydrous melts increases with depth, leading to efficient extraction of CO2 from the subducted slab at sub-arc conditions. Significant amounts of subducted carbon can be recycled back to the atmosphere via arc magmatism on relatively short time scales of less than 10 Myr. ACKNOWLEDGEMENTS The authors benefited from thoughtful and useful reviews by Ron Frost and Frieder Klein. We would like to thank Jim Beard for his editorial handling. James Tuff and Mike Shelley are acknowledged for early technical development on the gas chromatograph at RSES. We thank Greg Yaxley for comments, and Dave Clarke and Janet Hope for fruitful discussions and technical support for the gas chromatograph. Dave Clarke and Dean Scott are thanked for their technical support in the experimental laboratory. Malcolm Roberts and Bob Rapp are thanked for their assistance with the EPMA at CMCA-UWA and the SX100 at the RSES-ANU, respectively. FUNDING The authors acknowledge the facilities, and the scientific and technical assistance of the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy, Characterisation & Analysis, The University of Western Australia, a facility funded by the University, State and Commonwealth Governments. This work has been financially supported by the Australian Research Council. SUPPLEMENTARY DATA Supplementary data for this paper are available at Journal of Petrology online. REFERENCES Aeolus Lee C.-T. , Leeman W. P. , Canil D. , Li Z.-X. A. ( 2005 ). Similar V/Sc systematics in MORB and arc basalts: implications for the oxygen fugacities of their mantle source regions . Journal of Petrology 46 , 2313 – 2336 . Google Scholar CrossRef Search ADS Ague J. J. , Nicolescu S. ( 2014 ). Carbon dioxide released from subduction zones by fluid-mediated reactions . Nature Geosciences 7 , 355 – 360 . Google Scholar CrossRef Search ADS Akaishi M. , Yamaoka S. ( 2000 ). Crystallisation of diamond from C–O–H fluids under high-pressure and high-temperature conditions . Journal of Crystal Growth 209 , 999 – 1003 . Google Scholar CrossRef Search ADS Akaishi M. , Shaji Kumar M. D. , Kanda H. , Yamaoka S. ( 2000 ). Formation process of diamond from supercritical H2O–CO2 fluid under high pressure and high temperature conditions . Diamond and Related Materials 9 , 1945 – 1950 . Google Scholar CrossRef Search ADS Alt J. C. , Honnorez J. ( 1984 ). Alteration of the upper oceanic crust, DSDP site 417: mineralogy and chemistry . Contributions to Mineralogy and Petrology 87 , 149 – 169 . Google Scholar CrossRef Search ADS Alt J. C. , Teagle D. A. H. ( 1999 ). The uptake of carbon during alteration of ocean crust . Geochimica et Cosmochimica Acta 63 , 1527 – 1535 . Google Scholar CrossRef Search ADS Alt J. C. , Laverne C. , Vanko D. A. , Tartarotti P. , Teagle D. A. H. , Bach W. , Zuleger E. , Erzinger J. , Honnorez J. , Pezard P. A. ( 1996 ). Hydrothermal alteration of a section of upper oceanic crust in the eastern equatorial Pacific: a synthesis of results from site 504 (DSDP legs 69, 70, AND 83, and ODP legs 111, 137, 140, and 148). In: Alt, J. C., Kinoshita, H., Stokking, L. B. et al. (eds) Proceedings of the Ocean Drilling Program, Scientific Results , 148 . Ocean Drilling Program , pp. 417 – 434 . Alt J. C. , Schwarzenbach E. M. , Früh-Green G. L. , Shanks W. C. , Bernasconi S. M. , Garrido C. J. , Crispini L. , Gaggero L. , Padrón-Navarta J. A. , Marchesi C. ( 2013 ). The role of serpentinites in cycling of carbon and sulfur: seafloor serpentinization and subduction metamorphism . Lithos 178 , 40 – 54 . Google Scholar CrossRef Search ADS Arcay D. , Tric E. , Doin M. P. ( 2005 ). Numerical simulations of subduction zones: effect of slab dehydration on the mantle wedge dynamics . Physics of the Earth and Planetary Interiors 149 , 133 – 153 . Google Scholar CrossRef Search ADS Auzanneau E. , Schmidt M. W. , Vielzeuf D. , Connolly D. J. A. ( 2009 ). Titanium in phengite: a geobarometer for high temperature eclogites . Contributions to Mineralogy and Petrology 159 , 1 . Google Scholar CrossRef Search ADS Barker F. ( 1979 ). Trondhjemite: definition, environment, and hypotheses of origin. In: Barker F. (ed.) Trondhjemites, Dacites, and Related Rocks . Elsevier , pp. 1 – 12 . Becker H. ( 2000 ). Re–Os fractionation in eclogites and blueschists and the implications for recycling of oceanic crust into the mantle . Earth and Planetary Science Letters 177 , 287 – 300 . Google Scholar CrossRef Search ADS Becker K. , Sakai H. , Adamson A. C. , Alexandrovich J. , Alt J. C. , Anderson R. N. , Bideau D. , Gable R. , Herzig P. M. , Houghton S. , Ishizuka H. , Kawahata H. , Kinoshita H. , Langseth M. G. , Lovell M. A. , Malpas J. , Masuda H. , Merrill R. B. , Morin R. H. , Mottl M. J. , Pariso J. E. , Pezard P. , Phillips J. , Sparks J. , Uhlig S. ( 1989 ). Drilling deep into young oceanic crust, Hole 504B, Costa Rica Rift . Reviews of Geophysics 27 , 79 – 102 . Google Scholar CrossRef Search ADS Berner R. A. , Lasaga A. C. , Garrels R. M. ( 1983 ). The carbonate–silicate geochemical cycle and its effect on atmospheric carbon dioxide over the past 100 million years . American Journal of Science 283 , 641 – 683 . Google Scholar CrossRef Search ADS Cao Y. , Song S. G. , Niu Y. L. , Jung H. , Jin Z. M. ( 2011 ). Variation of mineral composition, fabric and oxygen fugacity from massive to foliated eclogites during exhumation of subducted ocean crust in the Norh Qilian suture zone, NW China . Journal of Metamorphic Geology 24 , 699 – 720 . Google Scholar CrossRef Search ADS Caciagli N. C. , Manning C. E. ( 2003 ). The solubility of calcite in water at 6–16 kbar and 500–800°C . Contributions to Mineralogy and Petrology 146 , 275 – 285 . Google Scholar CrossRef Search ADS Carter L. B. , Skora S. , Blundy J. D. , De Hoog J. C. M. , Elliott T. ( 2015 ). An experimental study of trace element fluxes from subducted oceanic crust . Journal of Petrology 56 , 1585 – 1606 . Google Scholar CrossRef Search ADS Chellappa R. S. , Somayazulu M. , Hemley R. J. ( 2009 ). Rhenium reactivity in H2O–O2 supercritical mixtures at high pressures . High Pressure Research 29 , 792 . Google Scholar CrossRef Search ADS Chepurov A. I. , Tomilenko A. A. , Zhimulev E. I. , Sonin V. M. , Chepurov A. A. , Kovyazin S. V. , Timina T. Y. , Surkov N. V. ( 2012 ). The conservation of an aqueous fluid in inclusions in minerals and their interstices at high pressures and temperatures during the decomposition of antigorite . Russian Geology and Geophysics 53 , 234 – 246 . Google Scholar CrossRef Search ADS Compagnoni R. , Rolfo F. ( 2003 ). Ultrahigh-pressure metamorphic rocks in the Western Alps. In: Carswell D. A. , Compagnoni R. (eds) Ultrahigh Pressure Metamorphism: EMU Notes in Mineralogy . Cambridge : Cambridge University press , pp. 206 – 243 . Connolly J. A. D. ( 2005 ). Computation of phase equilibria by linear programming: a tool for geodynamic modeling and its application to subduction zone decarbonation . Earth and Planetary Science Letters 236 , 524 – 541 . Google Scholar CrossRef Search ADS Dasgupta R. ( 2013 ). Ingassing, storage, and outgassing of terrestrial carbon through geologic time. In: Hazen R. M. , Jones A. P. , Baross J. A. (eds) Carbon in Earth. Mineralogical Society of America and Geochemical Society, Reviews in Mineralogy and Geochemistry 75 , 183 – 229 . Dasgupta R. , Hirschmann M. M. ( 2010 ). The deep carbon cycle and melting in the Earth’s interior . Earth and Planetary Science Letters 298 , 1 – 13 . Google Scholar CrossRef Search ADS Dasgupta R. , Hirschmann M. M. , Withers A. ( 2004 ). Deep global cycling of carbon constrained by the solidus of anhydrous, carbonated eclogite under upper mantle conditions . Earth and Planetary Science Letters 227 , 73 – 85 . Google Scholar CrossRef Search ADS de Leeuw G. A. M. , Hilton D. R. , Fischer T. P. , Walker J. A. ( 2007 ). The He–CO2 isotope and relative abundance characteristics of geothermal fluids in El Salvador and Honduras: new constraints on volatile mass balance of the Central American Volcanic Arc . Earth and Planetary Science Letters 258 , 132 – 146 . Google Scholar CrossRef Search ADS Edmond J. M. , Huh Y. ( 2003 ). Non-steady state carbonate recycling and implications for the evolution of atmospheric PCO2 . Earth and Planetary Science Letters 216 , 125 – 139 . Google Scholar CrossRef Search ADS Evans K. A. ( 2012 ). The redox budget of subduction zones . Earth-Science Reviews 113 , 11 – 32 . Google Scholar CrossRef Search ADS Evans K. A. , Tomkins A. G. ( 2011 ). The relationship between subduction zone redox budget and arc magma fertility . Earth and Planetary Science Letters 308 , 401 – 409 . Google Scholar CrossRef Search ADS Evans K. A. , Elburg M. A. , Kamenetsky V. S. ( 2012 ). Oxidation state of subarc mantle . Geology 40 , 783 – 786 . Google Scholar CrossRef Search ADS Fisk M. R. , Giovannoni S. J. , Thorseth I. H. ( 1998 ). Alteration of oceanic volcanic glass: textural evidence of microbial activity . Science 281 , 978 – 980 . Google Scholar CrossRef Search ADS Foley S. F. ( 2011 ). A reappraisal of redox melting in the Earth’s mantle as a function of tectonic setting and time . Journal of Petrology 52 , 1363 – 1391 . Google Scholar CrossRef Search ADS Frezzotti M. L. , Selverston J. , Sharp J. , Compagnoni R. ( 2011 ). Carbonate dissolution during subduction revealed by diamond-bearing rocks from the Alps . Nature Geoscience 4 , 703 – 706 . Google Scholar CrossRef Search ADS Galvez M. E. , Beyssac O. , Benzerara K. , Menguy N. , Bernard S. , Cox S. C. ( 2012 ). Micro- and nano-textural evidence of Ti(–Ca–Fe) mobility during fluid–rock interactions in carbonaceous lawsonite-bearing rocks from New Zealand . Contributions to Mineralogy and Petrology 164 , 895 – 914 . Google Scholar CrossRef Search ADS Gonzalez C. M. , Gorczyk W. , Gerya T. V. ( 2016 ). Decarbonation of subducting slabs: Insight from petrological-thermomechanical modeling . Gondwana Research 36 , 314 – 332 . Google Scholar CrossRef Search ADS Gorman P. J. , Kerrick D. M. , Connolly J. A. D. ( 2006 ). Modeling open system metamorphic decarbonation of subducting slabs . Geochemistry, Geophysics, Geosystems 7 , Q04007 . Google Scholar CrossRef Search ADS Gillis K. M. , Robinson P. T. ( 1990 ). Patterns and processes of alteration in the lavas and dykes of the Troodos Ophiolite, Cyprus . Journal of Geophysical Research 95 , 21523 – 21548 . Google Scholar CrossRef Search ADS Groppo C. , Castelli D. ( 2010 ). Prograde P–T evolution of a lawsonite eclogite from the Monviso Meta-ophiolite (Western Alps): dehydration and redox reactions during subduction of oceanic FeTi-oxide gabbro . Journal of Petrology 51 , 2489 – 2514 . Google Scholar CrossRef Search ADS Hammouda T. ( 2003 ). High-pressure melting of carbonated eclogite and experimental constraints on carbon recycling and storage in the mantle . Earth and Planetary Science Letters 214 , 357 – 368 . Google Scholar CrossRef Search ADS Harte B. ( 2010 ). Diamond formation in the deep mantle: the record of mineral inclusions and their distribution in relation to mantle dehydration zones . Mineralogical Magazine 74 , 189 – 215 . Google Scholar CrossRef Search ADS Hayes J. M. , Waldbauer J. R. ( 2006 ). The carbon cycle and associated redox processes through time . Philosophical Transactions of the Royal Society, Series B 361 , 931 – 950 . Google Scholar CrossRef Search ADS Hermann J. , Spandler C. ( 2008 ). Sediment melts at sub-arc depths: an experimental study . Journal of Petrology 49 , 717 – 740 . Google Scholar CrossRef Search ADS Hermann J. , Spandler C. , Hack A. , Korsakov A. V. ( 2006 ). Aqueous fluids and hydrous melts in high-pressure and ultra-high pressure rocks: implications for element transfer in subduction zones . Lithos 92 , 399 – 417 . Google Scholar CrossRef Search ADS Hermann J. , Troitzsch U. , Scott D. ( 2016 ). Experimental subsolidus phase relations in the system CaCO3–CaMg(CO3)2 up to 6·5 GPa and implications for subducted marbles . Contributions to Mineralogy and Petrology 171 , 84 . Google Scholar CrossRef Search ADS Hirschmann M. , Dasgupta R. ( 2009 ). The H/C ratios of Earth’s near-surface and deep reservoirs, and consequences for deep Earth volatile cycles . Chemical Geology 262 , 4 – 16 . Google Scholar CrossRef Search ADS Javoy M. , Pineau F. , Allègre C. J. ( 1982 ). Carbon geodynamic cycle . Nature 300 , 171 – 173 . Google Scholar CrossRef Search ADS Johnston F. K. B. , Turchyn A. V. , Edmonds M. ( 2011 ). Decarbonation efficiency in subduction zones: implications for warm Cretaceous climates . Earth and Planetary Science Letters 303 , 143 – 152 . Google Scholar CrossRef Search ADS Kelemen P. B. , Manning C. E. ( 2015 ). Reevaluating carbon fluxes in subduction zones, what goes down, mostly comes up . Proceedings of the National Academy of Sciences of the USA 112 , E3997 – E4006 . Google Scholar CrossRef Search ADS Kelley K. A. , Plank T. , Ludden J. , Staudigel H. ( 2003 ). Composition of altered oceanic crust at ODP Sites 801 and 1149 . Geochemistry, Geophysics, Geosystems 4 , 8910 . Google Scholar CrossRef Search ADS Kerrick D. M. , Connolly J. A. D. ( 1998 ). Subduction of ophicarbonates and recycling of CO2 and H2O . Geology 26 , 375 – 378 . Google Scholar CrossRef Search ADS Kerrick D. M. , Connolly J. A. D. ( 2001 ). Metamorphic devolatilization of subducted oceanic metabasalts: implications for seismicity, arc magmatism and volatile recycling . Earth and Planetary Science Letters 189 , 19 – 29 . Google Scholar CrossRef Search ADS Kessel R. , Ulmer P. , Pettke T. , Schmidt M. W. , Thompson A. B. ( 2005 ). The water–basalt system at 4 to 6 GPa: Phase relations and second critical endpoint in a K-free eclogite at 700 to 1400°C . Earth and Planetary Science Letters 237 , 873 – 892 . Google Scholar CrossRef Search ADS Kiseeva E. S. , Yaxley G. M. , Hermann J. , Litasov K. D. , Rosenthal A. , Kamenetsky V. S. ( 2012 ). An experimental study of carbonated eclogite at 3.5–5.5 GPa—implications for silicate and carbonate metasomatism in the cratonic mantle . Journal of Petrology 53 , 727 – 759 . Google Scholar CrossRef Search ADS Kono Y. , Kenney-Benson C. , Hummer D. , Ohfuji H. , Park C. , Shen G. , Wang Y. , Kavner A. , Manning C. E. ( 2014 ). Ultralow viscosity of carbonate melts at high pressures . Nature Communications 5 , 5091 . Google Scholar CrossRef Search ADS Lassiter J. C. ( 2003 ). Rhenium volatility in subaerial lavas: constraints from subaerial and submarine portions of the HSDP-2 Mauna Kea drillcore . Earth and Planetary Science Letters 214 , 311 . Google Scholar CrossRef Search ADS Martin L. H. J. , Schmidt M. W. , Mattsson H. B. , Guenther D. ( 2013 ). Element partitioning between immiscible carbonatite and silicate melts for dry and H2O-bearing systems at 1–3 GPa . Journal of Petrology 54 , 2301 – 2338 . Google Scholar CrossRef Search ADS Marty B. , Tolstikhin I. N. ( 1998 ). CO2 fluxes from mid-ocean ridges, arcs and plumes . Chemical Geology 145 , 233 – 248 . Google Scholar CrossRef Search ADS Mattison C. G. , Tsujimori T. , Zhang R. Y. , Liou L. G. ( 2004 ). Epidote-rich talc-kyanite-phengite eclogites, Sulu terranes, eastern China: P-T-fO2 estimates and the significance of the epidote-talc assemblage in eclogite . American Mineralogist 89 , 1772 – 1783 . Google Scholar CrossRef Search ADS McCubbin F. M. , Sverjensky D. A. , Steele A. , Mysen B. O. ( 2014 ). In-situ characterization of oxalic acid breakdown at elevated P and T: implications for organic C–O–H fluid sources in petrologic experiments . American Mineralogist 99 , 2258 – 2271 . Google Scholar CrossRef Search ADS Molina J. F. , Poli S. ( 2000 ). Carbonate stability and fluid composition in subducted oceanic crust: an experimental study on H2O–CO2-bearing basalts . Earth and Planetary Science Letters 176 , 295 – 310 . Google Scholar CrossRef Search ADS Morgan G. B. , Chou I.-M. , Pasteris J. D. ( 1992 ). Speciation in experimental C–O–H fluids produced by the thermal dissociation of oxalic acid dihydrate . Geochimica et Cosmochimica Acta 56 , 281 – 294 . Google Scholar CrossRef Search ADS Morimoto N. ( 1988 ). Nomenclature of pyroxenes . Mineralogy and Petrology 39 , 55 – 76 . Google Scholar CrossRef Search ADS Morris J. D. , Leeman W. P. , Tera F. ( 1990 ). The subducted component in island arc lavas: constraints from Be isotopes and B–Be systematics . Nature 344 , 31 – 36 . Google Scholar CrossRef Search ADS Parkinson I. J. , Arculus R. J. ( 1999 ). The redox state of subduction zones: insights from arc-peridotites . Chemical Geology 160 , 409 – 423 . Google Scholar CrossRef Search ADS Pawley A. R. , Holloway J. R. ( 1993 ). Water sources for subduction zone volcanism: new experimental constraints . Science 260 , 664 – 667 . Google Scholar CrossRef Search ADS Peacock S. M. ( 1996 ). Thermal and petrologic structure of subductions zones. In: Bebout G. E. , Scholl D. W. , Kirby S. H. , Platt J. P. (eds) Subduction: Top to Bottom. American Geophysical Union, Geophysical Monograph 96 , 119 – 133 . Plank T. , Cooper L. B. , Manning C. E. ( 2009 ). Emerging geothermometers for estimating slab surface temperatures . Nature Geosciences 2 , 611 – 615 . Google Scholar CrossRef Search ADS Poli S. ( 2015 ). Carbon mobilized at shallow depths in subduction zones by carbonatitic liquids . Nature Geoscience 8 , 633 – 636 . Google Scholar CrossRef Search ADS Poli S. , Schmidt M. W. ( 1995 ). H2O transport and release in subduction zones: experimental constraints on basaltic and andesitic system . Journal of Geophysical Research 100 , 22299 – 222314 . Google Scholar CrossRef Search ADS Poli S. , Franzolin E. , Fumagalli P. , Crottini A. ( 2009 ). The transport of carbon and hydrogen in subducted oceanic crust: an experimental study to 5 GPa . Earth and Planetary Science Letters 278 , 350 – 360 . Google Scholar CrossRef Search ADS Rohrbach A. , Schmidt M. W. ( 2011 ). Redox freezing and melting in the Earth’s deep mantle resulting from carbon–iron redox coupling . Nature 472 , 209 – 212 . Google Scholar CrossRef Search ADS Rowe M. C. , Kent A. J. R. , Nielsen R. L. ( 2009 ). Subduction influence on oxygen fugacity and trace and volatile elements in basalts across the Cascade volcanic arc . Journal of Petrology 50 , 61 – 91 . Google Scholar CrossRef Search ADS Sano Y. , Marty B. ( 1995 ). Origin of carbon in fumarolic gas from island arcs . Chemical Geology 119 , 265 – 274 . Google Scholar CrossRef Search ADS Schmidt M. W. , Poli S. ( 1998 ). Experimentally based water budgets for dehydrating slabs and consequences for arc magma generation . Earth and Planetary Science Letters 163 , 361 – 379 . Google Scholar CrossRef Search ADS Schmidt M. W. , Vielzeuf D. , Auzanneau E. ( 2004 ). Melting and dissolution of subducting crust: the key role of white micas . Earth and Planetary Science Letters 228 , 65 – 84 . Google Scholar CrossRef Search ADS Sciurto P. F. , Ottonello G. ( 1995 ). Water–rock interaction on Zabargad Island, Red Sea—a case study: I. Application of the concept of local equilibrium . Geochimica et Cosmochimica Acta 59 , 2187 – 2206 . Google Scholar CrossRef Search ADS Shaw A. M. , Hilton D. R. , Fischer T. P. , Walker J. A. , Alvarado G. E. ( 2003 ). Contrasting He–C relationships in Nicaragua and Costa Rica: insights into C cycling through subduction zones . Earth and Planetary Science Letters 214 , 499 – 513 . Google Scholar CrossRef Search ADS Simakin A. G. , Salova T. P. , Bondarenko G. V. ( 2012 ). Experimental study of magmatic melt oxidation by CO2 . Petrology 20 , 593 – 606 . Google Scholar CrossRef Search ADS Skora S. , Blundy J. D. , Brooker R. A. , Green E. C. R. , de Hoog J. C. M. , Connolly J. A. D. ( 2015 ). Hydrous phase relations and trace element partitioning behaviour in calcareous sediments at subduction-zone conditions . Journal of Petrology 56 , 953 – 980 . Google Scholar CrossRef Search ADS Smit M. A. , Bröcker M. , Scherer E. E. ( 2008 ). Aragonite and magnesite in eclogites from the Jæren nappe, SW Norway: disequilibrium in the system CaCO3–MgCO3 and petrological implications . Journal of Metamorphic Geology 26 , 959 – 979 . Google Scholar CrossRef Search ADS Spandler C. , Pettke T. , Rubatto D. ( 2011 ). Internal and external fluid sources for eclogite-facies veins in the Monviso Meta-ophiolite, Western Alps: implications for fluid flow in subduction zone . Journal of Petrology 52 , 1207 – 1236 . Google Scholar CrossRef Search ADS Staudigel H. , Hart S. R. ( 1983 ). Alteration of basaltic glass: Mechanisms and significance for the oceanic crust-seawater budget . Geochimica et Cosmochimica Acta 47 , 337 – 350 . Google Scholar CrossRef Search ADS Stakes D. S. , O’Neil J. R. ( 1982 ). Mineralogy and stable isotope geochemistry of hydrothermally altered oceanic rocks . Earth and Planetary Science Letters 57 , 285 – 304 . Google Scholar CrossRef Search ADS Staudigel H. ( 2003 ). Hydrothermal alteration processes in the oceanic crust. In: Rudnick R. L. (ed.) The Crust . Elsevier , pp. 511 – 535 . Staudigel H. , Hart S. R. , Richardson S. H. ( 1981 ). Alteration of the oceanic crust: processes and timing . Earth and Planetary Science Letters 52 , 311 – 327 . Google Scholar CrossRef Search ADS Staudigel H. , Hart S. R. , Schmincke H.-U. , Smith B. M. ( 1989 ). Cretaceous ocean crust at DSDP Sites 417 and 418: Carbon uptake from weathering versus loss by magmatic outgassing . Geochimica et Cosmochimica Acta 53 , 3091 – 3094 . Google Scholar CrossRef Search ADS Staudigel H. , Plank T. , White B. , Schmincke H.-U. ( 1996 ). Geochemical fluxes during seafloor alteration of the basaltic upper oceanic crust: DSDP Sites 417 and 418. In: Bebout E. , Scholl W. , Kirby H. , Platt P. (eds) Subduction Top to Bottom. American Geophysical Union, Geophysical Monograph 96 , 19 – 38 . Sverjensky D. A. , Stagno V. , Huang F. ( 2014 ). Important role for organic carbon in subduction-zone fluids in the deep carbon cycle . Nature Geoscience 7 , 909 – 913 . Google Scholar CrossRef Search ADS Syracuse E. M. , van Keken P. E. , Abers G. A. ( 2010 ). The global range of subduction zone thermal models . Physics of the Earth and Planetary Interiors 183 , 73 – 90 . Google Scholar CrossRef Search ADS Taylor W. R. , Foley S. F. ( 1989 ). Improved oxygen-buffering techniques for C–O–H fluid-saturated experiments at high pressure . Journal of Geophysical Research 94 , 4146 – 4158 . Google Scholar CrossRef Search ADS Thomsen T. B. , Schmidt M. W. ( 2008 ). Melting of carbonated pelites at 2.5-5.0 GPa, silicate-carbonatite liquid immiscibility, and potassium-carbon metasomatism of the mantle . Earth and Planetary Science Letters 267 , 17 – 31 . Google Scholar CrossRef Search ADS Tiraboschi C. , Tumiati S. , Recchia S. , Miozzi F. , Poli S. ( 2016 ). Quantitative analysis of COH fluids synthesized at HP–HT conditions: an optimized methodology to measure volatiles in experimental capsules . Geofluids 16 , 841 – 855 . Google Scholar CrossRef Search ADS Tsuno K. , Dasgupta R. , Danielson L. , Righter K. ( 2012 ). Flux of carbonate melt from deeply subducted pelitic sediments: Geophysical and geochemical implications for the source of Central American volcanic arc . Geophysical Research Letters 39 , L16307 . Google Scholar CrossRef Search ADS Tumiati S. , Godard G. , Martin S. , Malaspina N. , Poli S. ( 2015 ). Ultra-oxidized rocks in subduction melanges? Decoupling between oxygen fugacity and oxygen availability in a Mn-rich metasomatic environment . Lithos 226 , 116 – 130 . Google Scholar CrossRef Search ADS van Keken P. E. , Kiefer B. , Peacock S. M. ( 2002 ). High-resolution models of subduction zones: Implications for mineral dehydration reactions and the transport of water into the deep mantle . Geochemistry, Geophysics, Geosystems 3 , 1–20 . Google Scholar CrossRef Search ADS Varekamp J. C. , Kreulen R. , Poorter R. P. E. , Van Bergen M. J. ( 1992 ). Carbon sources in arc volcanism with implications for the carbon cycle . Terra Nova 140 , 217 – 240 . Wallace P. J. ( 2005 ). Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data . Journal of Volcanology and Geothermal Research 140 , 217 – 240 . Google Scholar CrossRef Search ADS Wenzel T. , Baumgartner L. P. , Brogmann G. E. , Konnikov E. G. , Kislov E. V. ( 2002 ). Partial melting and assimilation of dolomitic xenoliths by mafic magma: the Ioko-Dovyren intrusion (North Baikal region, Russia) . Journal of Petrology 43 , 2049 – 2074 . Google Scholar CrossRef Search ADS Wyllie P. J. , Tuttle O. F. ( 1959 ). Melting of calcite in the presence of water . American Mineralogist 44 , 453 – 461 . Yaxley G. M. , Green D. H. ( 1994 ). Experimental demonstration of refractory carbonate-bearing eclogite and siliceous melt in the subduction regime . Earth and Planetary Science Letters 128 , 313 – 325 . Google Scholar CrossRef Search ADS Zhang R. Y. , Liou J. G. , Ernst W. G. ( 2009 ). The Dabie–Sulu continental collision zone: a comprehensive review . Gondwana Research 16 , 1 – 26 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Petrology Oxford University Press

Experimental Phase Relations in Altered Oceanic Crust: Implications for Carbon Recycling at Subduction Zones

Loading next page...
 
/lp/ou_press/experimental-phase-relations-in-altered-oceanic-crust-implications-for-FVqGccDXMY
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0022-3530
eISSN
1460-2415
D.O.I.
10.1093/petrology/egy031
Publisher site
See Article on Publisher Site

Abstract

ABSTRACT The phase relationships in altered mafic oceanic crust (K2O, CO2 and H2O bearing) have been investigated to constrain and quantify the processes of carbon transfer from the slab to the mantle wedge at subduction zones. We report experiments at 2·5–4·5 GPa and 700–900°C in which gas in the experimental charges is analysed by gas chromatography to constrain the volatile composition of the aqueous fluid or hydrous melt. A phengite-bearing epidote-eclogite with dolomite and/or magnesite is stable at subsolidus conditions. The wet solidus at fO2 of NNO (nickel–nickel oxide) was found between 700 and 750°C at 2·5–3·5 GPa and 800–850°C at 4·5 GPa, similar to the CO2-free systems. This observation indicates a low amount of CO2 in the aqueous fluid phase at the wet solidus, in agreement with a measured X(CO2) of 0·059 ± 0·003 at 3 GPa, 700°C and 0·038 ± 0·003 at 3·5 GPa, 700°C. Experiments performed at higher fO2, using either an oxidized starting material or a Re–ReO2 buffer, resulted in a shift of the solidus to higher temperatures. Higher fO2 results in a higher X(CO2) in the aqueous fluid and the reduced water activity leads to a shift in the solidus to higher temperatures. Above the solidus, both textural observations and analyses of the gas enclosed in the capsule suggest that CO2 solubility in the silicate melt increases with increasing P–T conditions. At 4·5 GPa, more than 70% of carbonate at 850°C and 100% at 900°C was dissolved in the hydrous silicate melt. From textural observation it is not clear whether the high carbon content of the melt is related to an increased solubility of carbonate in the hydrous silicate melt or reflects an immiscible carbonatite melt. In any case, partial melting of altered oceanic crust at moderately oxidizing conditions (NNO) and pressures >4 GPa provides an efficient means for the transfer of carbon from the slab to the mantle wedge in intermediate to hot subduction zones. Significant amounts of subducted carbon can thus be brought back to the atmosphere via arc magmatism on relatively short time scales of less than 10 Myr. INTRODUCTION Understanding the deep carbon cycle is important for constraining the evolution of climate over geological time scales (∼100 Myr) (Berner et al., 1983; Edmond & Huh, 2003), the formation of diamonds in the mantle (Harte, 2010; Sverjensky et al., 2014), as well as many processes controlling the chemical evolution of the Earth’s interior (Dasgupta & Hirschmann, 2010; Dasgupta, 2013). Subduction zones play a significant role in the deep carbon cycle as they are the only geological system on Earth that allows carbon to be transported into the deep mantle, which, in turn, is degassed through magmatism (Javoy et al., 1982; Marty & Tolstikhin, 1998). However, whether carbon is effectively recycled towards the atmosphere through arc magmatism located above subduction zones, or buried into the deep mantle, remains controversial, as are the processes involved in carbon mobility within the slab (Fig. 1). The geochemistry of volatiles (mainly the CO2/3He ratio) and basalts (melt inclusions or glass) from arc volcanoes suggests that most of the carbon emitted originates from the subducting slab (Sano & Marty, 1995; Marty & Tolstikhin, 1998; Wallace, 2005). Estimates of the recycling efficiency of carbon at subduction zones are based on the comparison between the carbon fluxes entering subduction zones and outgassing at the surface (Kelemen & Manning, 2015). These estimates differ between studies from 30% to nearly 100% (Varekamp et al., 1992; Marty & Tolstikhin, 1998; Dasgupta & Hirschmann, 2010; Johnston et al., 2011; Kelemen & Manning, 2015) but can be as low as 18–20% (Shaw et al., 2003; de Leeuw et al., 2007; Johnston et al., 2011) (Fig. 1). One source of variability comes from the carbon flux entering the subduction system, which is related to the lithological and chemical composition of the slab, especially the ratio between sediments and altered oceanic crust, the degree of alteration of the oceanic crust and the extent of serpentinization in the subducting lithosphere (Sciurto & Ottonello, 1995; Kerrick & Connolly, 1998; Dasgupta & Hirschmann, 2010; Evans, 2012; Alt et al., 2013). Nevertheless, estimates of carbon fluxes at subduction zones show that the altered oceanic crust is, overall, one of the main hosts of carbon (Staudigel et al., 1989; Alt & Teagle, 1999; Hayes & Waldbauer, 2006; Dasgupta & Hirschmann, 2010; Evans, 2012; Alt et al., 2013; Dasgupta, 2013; Kelemen & Manning, 2015) (Fig. 1). In contrast to geochemical studies, thermodynamic calculations and experimental studies on the stability of carbonate in the subducting altered oceanic crust have mainly yielded contrasting results. Earlier experimental studies conducted on altered oceanic crust suggested that decarbonation mostly occurs at fore-arc depths (Molina & Poli, 2000; Poli et al., 2009) or at much higher temperature conditions (in anhydrous eclogite T > 1050°C and P > 3·5 GPa; Kiseeva et al., 2012). In sediments, decarbonation occurs through the formation of carbonatite at melting conditions of P > 3 GPa, T > 1050°C (Tsuno et al., 2012; Thomsen & Schmidt, 2008), and in carbonate-rich sediments at P = 3 GPa, at T > 1000°C (Skora et al., 2015), conditions that are consistent with only the hottest geotherms of subduction zones or pressures beyond sub-arc depths. At sub-arc depths, experiments (Yaxley & Green, 1994; Molina & Poli, 2000; Hammouda, 2003; Dasgupta et al., 2004; Poli et al., 2009; Skora et al., 2015) and thermodynamic modelling (Kerrick & Connolly, 2001; Gorman et al., 2006; Gonzalez et al., 2016) have shown that carbon is mainly trapped in carbonates or graphite and thus carried into the deep mantle. Fluids released from the subducting altered oceanic crust at sub-arc depths are then expected to have low C/H ratios. Recent work by Poli (2015) has shown that altered plutonic rocks forming the lower part of the oceanic crust experience partial melting at relatively low pressure–temperature conditions corresponding to sub-arc depths (900°C and 3·8–4·2 GPa), with the occurrence of both carbonatitic and silicate melts. This provides an alternative process for C release from the subducted slab. So far, no experiments on altered basaltic composition have been conducted to evaluate whether or not C can be released by melting of altered basalts. Fig. 1. View largeDownload slide Schematic view of the deep carbon cycle in subduction zones. The numbers in black represent the estimation of the yearly fluxes of C (in Tera mols) in and out of subduction zones (Kerrick & Connolly, 2001; Hayes & Waldbauer, 2006; Dasgupta & Hirschmann, 2010; Alt et al., 2013; Dasgupta, 2013; Kelemen & Manning, 2015). Our experimental work shows that along a cold geotherm, 70–90 wt % of C is trapped in residual carbonates. Conversely, in intermediate and hot geotherms, 25–100 wt % stored in the altered oceanic crust can leave the slab via partial melting at sub-arc depth. At low P–T conditions, fluids are produced by dehydration of the subducting lithologies and C is trapped in carbonates (1) (this study). At super-solidus conditions, either a Si-rich melt (2) (this study) or both Si and C melts (3) (this study; Poli, 2015) are produced, allowing for efficient decarbonation of the subducting altered oceanic crust. Schematic representation of textural and solid versus liquid phase relationships at P–T and 1 atm is given on the right-hand side of the figure. Fig. 1. View largeDownload slide Schematic view of the deep carbon cycle in subduction zones. The numbers in black represent the estimation of the yearly fluxes of C (in Tera mols) in and out of subduction zones (Kerrick & Connolly, 2001; Hayes & Waldbauer, 2006; Dasgupta & Hirschmann, 2010; Alt et al., 2013; Dasgupta, 2013; Kelemen & Manning, 2015). Our experimental work shows that along a cold geotherm, 70–90 wt % of C is trapped in residual carbonates. Conversely, in intermediate and hot geotherms, 25–100 wt % stored in the altered oceanic crust can leave the slab via partial melting at sub-arc depth. At low P–T conditions, fluids are produced by dehydration of the subducting lithologies and C is trapped in carbonates (1) (this study). At super-solidus conditions, either a Si-rich melt (2) (this study) or both Si and C melts (3) (this study; Poli, 2015) are produced, allowing for efficient decarbonation of the subducting altered oceanic crust. Schematic representation of textural and solid versus liquid phase relationships at P–T and 1 atm is given on the right-hand side of the figure. When the oceanic crust is altered by seawater, it gains H2O and CO2, but also some K2O (Kelley et al., 2003; Staudigel, 2003), all elements potentially affecting the melting temperature in mafic rocks (Schmidt et al., 2004; Carter et al., 2015). Altered oceanic crust (AOC) is located at the top of the subducting slab, where the temperatures are the hottest (i.e. van Keken et al., 2002) and thus fluid fluxed melting might be possible (Arcay et al., 2005; Syracuse et al., 2010), as investigated in this contribution. Experimental studies are hampered additionally by the difficulty of estimating the volatile composition of the fluid or melt in equilibrium with the solid assemblages composing the metamorphosed altered oceanic crust. Although the volatile composition can theoretically be obtained indirectly from mass-balance calculations, in many cases this method is not capable of resolving small differences in residual carbonates, which inevitably have a large influence on the calculated CO2 content of the fluid phase. To circumvent this problem, analytical techniques allowing the analysis of the gas present in the experimental charge have been developed, including Raman spectrometry (Morgan et al., 1992; Akaishi & Yamaoka, 2000) and mass spectrometry coupled with gas chromatography (Taylor & Foley, 1989; Akaishi et al., 2000; Chepurov et al., 2012; Tiraboschi et al., 2016). In this contribution, we experimentally explore the effect of H2O, CO2 and K2O added during alteration processes on the phase relationships in the AOC and estimate the composition of the gas present in the capsule after the experiment with a gas chromatograph equipped with a thermal conductivity detector (GC-TCD). Our results show the following: (1) at subsolidus conditions, the fluid is water-rich, consistent with previous studies (Kerrick & Connolly, 2001; Poli et al., 2009); (2) the altered oceanic crust partially melts at sub-arc depths conditions; (3) carbonate solubility increases with pressure in the resultant melt, providing a mechanism to transfer carbon from the subducted slab to the mantle wedge. ANALYTICAL AND EXPERIMENTAL METHODS The altered oceanic crust The AOC composition and modes of alteration are mostly inferred from ophiolites (e.g. Gillis & Robinson, 1990) and direct sampling by deep basement drillings (e.g. Staudigel et al., 1981; Alt & Honnorez, 1984; Staudigel, 2003, and references therein). Although deep drilling through young oceanic crust such as hole 504B (Costa Rica Rift, 5·9 Ma) provides information about the early hydrothermal alteration processes (Becker et al., 1989), typical reference composition and style of alteration for the AOC comes from the Ocean Drilling Program (ODP) in older oceanic crust (>100 Ma, such as sites 417/418, 801 and 1149) (Staudigel, 2003). As the mean age for the subducted oceanic crust is older than 60 Ma, these sites provide the AOC end-member component used in discussions about the geochemical processes occurring in subduction zones (e.g. Staudigel, 2003). In the following, we summarize the effect of the alteration on the CO2, H2O and K2O content of the oceanic crust. The oceanic crust experiences important alteration that varies in time and space from the initial cooling at the contact with seawater to its evolution as it moves away from the axial zone (i.e. Staudigel et al., 1981; Fisk et al., 1998; Alt & Teagle, 1999; Staudigel, 2003). In the first 300–600 m, secondary minerals include chlorite, celadonite, saponite, calcite and different Fe-oxides or hydroxides crystallizing in voids or interstices, veins, alteration haloes and fracture fillings (Stakes & O’Neil, 1982; Alt & Honnorez, 1984; Becker et al., 1989; Alt et al., 1996; Alt & Teagle, 1999; Staudigel, 2003). This alteration is the result of complex, subsequent reactions leading to an important uptake of CO2, H2O and K2O (Staudigel & Hart, 1983; Alt & Honnorez, 1984; Staudigel, 2003). In contrast, the deeper parts of the oceanic crust (deeper extrusive and sheeted dykes) experience a more pervasive alteration occurring at higher temperatures, resulting in the crystallization of greenschist- to amphibolite-facies minerals in replacement of primary magmatic phases (Alt & Honnorez, 1984). These secondary minerals, which include pumpellyite, chlorite and calcite amongst others (Staudigel, 2003), reflect enrichment in H2O and CO2 compared with unaltered oceanic crust. Staudigel (2003) noted that the carbon intake of young oceanic crust (sites 504B and 332B) is variable, albeit overall lower than that of mature oceanic crust. Young oceanic crust, such as at site 504B, can thus show different ratios between CO2, K2O and H2O (Alt et al., 1996). An average composition of the altered oceanic crust has been estimated from drill cores, which provide sections of the upper parts of the altered oceanic crust (Alt et al., 1996; Staudigel et al., 1996). The composition of the ‘supercomposite’ was estimated from the mixing of different lithologies in typical proportions found in the altered oceanic crust (Staudigel et al., 1989) and is given in Table 1. Table 1: Composition of the starting materials Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 * Staudigel et al. (1989). In SM3O, all iron is oxidized (see text for details). Table 1: Composition of the starting materials Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 Supercomposite* SM2 SM3 SM3O SM2d SiO2 47·2 46·26 47·60 47·76 48·41 Al2O3 15·1 13·94 13·83 13·86 14·59 TiO2 1·15 1·10 1·31 1·02 1·15 MgO 6·50 6·06 6·26 6·25 6·34 FeOT 8·80 8·15 7·90 7·99 8·53 CaO 12·6 11·47 11·23 11·22 12·01 Na2O 2·02 1·89 1·91 1·91 1·98 K2O 0·94 0·88 0·89 0·89 0·92 CO2 2·97 2·77 2·82 2·83 2·90 H2O 2·70 7·48 6·25 6·26 3·18 Total 100 100 100 100 100 * Staudigel et al. (1989). In SM3O, all iron is oxidized (see text for details). Starting material The compositions of our starting materials are close to the ‘supercomposite’ (Staudigel et al., 1989). However, the K2O content was increased to ∼0·9 wt % to ease phengite identification in experimental runs and to be consistent with more recent estimations of the altered oceanic crust composition (e.g. Kelley et al., 2003). Additional water was added (∼7 wt %) to allow investigation of CO2 release during influx of aqueous fluids from deeper levels of the subducted crust, which is considered the main process for mobilizing slab-derived carbon, major and trace elements (Connolly, 2005; Hermann et al., 2006; Hermann & Spandler, 2008; Plank et al., 2009), and simulate fluid-fluxed melting. Although the altered oceanic crust is likely to experience some decarbonation at lower pressures (<3 GPa; Molina & Poli, 2000; Ague & Nicolescu, 2014), the CO2 content of the supercomposite was kept unchanged to represent an end-member composition for the altered oceanic crust. All starting materials used in the experiments have similar major element compositions. The starting materials SM2 and SM3 were made of synthetic powders mixed in two steps. In the first step, SiO2 as a silica gel, TiO2, MgO together with CaO, K2O and Na2O as carbonates were mixed in an agate mortar. This mix was then pressed into a pellet and fully decarbonated by heating at 800–1000°C in a furnace for 12 h. In a second step, Fe2SiO4, Al(OH)3 and CaCO3 were added to the first mix in the proportions required to obtain the targeted composition. The starting material SM3O was prepared under the same conditions as SM2 and SM3, except that Fe2SiO4 was added before heating at 800–1000°C, ensuring the oxidation of all Fe2+ to Fe3+. The starting material SM2d was obtained by partial dehydration of SM2 in a furnace. The dehydration of Al(OH)3, which is the source of water in the starting materials, was calibrated as a function of time to retrieve the amount of water remaining in SM2d. All three starting materials SM2, SM3 and SM3O were mixed with Li2B2O7 in the proportion 1:4 and then fused to a glass. These glasses were analysed for major elements using a Cameca SX100 operating at 15 kV and 10 nA with a defocused beam at the Research School of Earth Sciences at the ANU. The CO2 and H2O contents of the starting material were calculated from the weight per cent of CaO and Al2O3 measured in the glass and masses of CaCO3 and Al(OH)3 powders added to the starting material. The composition of the starting materials and recalculated CO2 and H2O contents are given in Table 1. Experimental setup Experiments AOC28, 25, 29, 30, 12 and Re4 were designed to estimate the effect of oxygen fugacity on C-bearing phase stability and the phase relations near the solidus (Table 2). Experiments AOC28, 25 and 29 are composed of different proportions (20, 50, 75%) of oxidized (SM3O) and reduced (SM2) starting materials (Table 1), and were run at 3 GPa, 750°C, which is just above the solidus. Experiment AOC30 contained the most oxidized starting material (SM3O) alone. In non-buffered experiments, the assembly (see below) imposes an oxygen fugacity close to NiNiO (NNO; Hermann & Spandler, 2008). The presence of an Fe–Ti oxide, pertaining to the magnetite–ulvöspinel solid-solution in experiment AOC30, suggests that the other experiments made of a mix between reduced and oxidized starting materials have fO2 intermediate between NNO and Magnetite–Hematite (MH) equilibria. Experiments AOC12 and Re4 were buffered by adding a mix of Re–ReO2 at the bottom and top of the capsule. The Re and ReO2 powders were mixed in the proportion 1:5. The Re–ReO2 and SM2 powders were mixed in the proportion 1:1. Experiment Re4 was designed for gas analysis by a gas chromatograph equipped with a thermal conductivity detector (GC-TCD) and thus contained a larger mass of starting material and Re–ReO2 mix (76·7 mg against 36·2 mg in AOC12). The experiments analysed by GC-TCD (AOC40–43, 46 and D850-45) were run with ∼74–95 mg of 0·7SM3 + 0·3SM3O starting material loaded in 3·5 mm inner diameter Au capsules. All the other experiments were run with 2·3 mm inner diameter gold capsules loaded with ∼15–20 mg of SM2 and SM2d starting material. Table 2: Starting materials, P, T and run product in experiments SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – 1 SM, Starting material. 2 30SM3O: Mix of 30%SM3O+70%SM2. 3 Ca-features includes Ca-rich pools and skeletal crystals (DC: dendritic crystals, S: spherules). 4 Fe, Ti: Fe, Ti-rich oxides. 5 Phase proportions calculated with volatile melt composition obtained from GC-TCD analyses (see Table 5) * Phases were identified only. Table 2: Starting materials, P, T and run product in experiments SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – SM1 Name P T Garnet Omphacite Phengite Epidote Lawsonite Kyanite SiO2 Rutile Melt Ca-rich features3 Mg- calcite Dolomite Magnesite Aragonite Fe-Ti- oxide4 fO2 experiments 20SM3O AOC28 3 750 0·25 0·33 0·06 0·09 – – 0·01 0·01 0·19 – – 0·06 – – – 50SM3O AOC25 3 750 0·24 0·35 0·06 0·07 – – 0·01 0·01 0·20 – – 0·06 – – – 75SM3O AOC29 3 750 0·25 0·37 0·07 0·08 – – 0·02 0·01 0·15 – – 0·06 – – – 100SM3O AOC30 3 750 0·04 0·37 0·08 0·26 – 0·03 0·10 0·00 – – – 0·05 – 0·01 SM2+ReRe2O AOC12 3 750 0·17 0·40 0·08 0·08 – 0·05 0·08 0·01 – – – 0·07 – – – SM2+ReRe2O Re4 3 750 * * * * * * * * Other experiments SM2 AOC10 2·5 700 0·20 0·34 0·09 0·13 – – 0·10 0·01 <0·01 – – 0·04 0·02 – – SM2 AOC9 2·5 750 0·29 0·36 0·08 <0·01 – – – 0·01 0·20 – – 0·06 – – – SM2 AOC6 2·5 800 0·35 0·24 – <0·01 – – – 0·00 0·34 – 0·06 – – – – SM2 AOC7 2·5 850 0·37 0·22 – <0·01 – – – 0·00 0·35 0·01 (DC) 0·05 – – – – 30SM3O2 AOC41 3 700 0·21 0·40 0·09 0·11 – – 0·12 0·01 – – – 0·01 0·05 – – SM2 AOC20 3 700 0·25 0·42 0·08 0·09 – 0·02 0·06 0·02 – – – 0·03 0·03 – – SM2 AOC13 3 750 0·26 0·36 0·05 0·05 – – 0·02 0·01 0·18 – – 0·06 – – – 30SM3O AOC46 3 775 0·33 0·35 0·05 <0·01 – – 0·00 0·01 0·20 – 0·06 – – – With melt composition from GC5 0·24 0·41 0·08 – – – 0·00 0·01 0·22 – 0·06 – – – AOC42 3 775 * * * – – – – * – – * – – – SM2 AOC15 3 800 0·34 0·26 0·03 – – – – 0·01 0·31 – 0·06 – – – – SM2 AOC21 3 850 0·35 0·23 – – – – – 0·01 0·34 0·06 (S) – – – – 30SM3O AOC40 3·5 700 0·24 0·42 0·09 0·01 – 0·03 0·09 0·01 – – – 0·06 0·00 – – SM2 AOC11 3·5 700 0·21 0·46 0·09 <0·01 – 0·05 0·06 0·01 – – – 0·04 0·01 – – SM2 AOC8 3·5 750 0·34 0·34 0·04 0·02 – – 0·00 0·01 0·19 – – 0·06 – – – 30SM3O AOC43 3·5 775 0·32 0·33 0·05 <0·01 – – <0·01 0·01 0·24 – 0·04 – – – – With melt composition from GC 0·33 0·30 0·06 0·03 0·07 <0·01 0·15 0·04 – – – – SM2 AOC5 3·5 800 0·37 0·25 0·03 <0·01 – – – 0·01 0·27 – 0·06 – – – – 30SM3O D700-45 4·5 700 0·26 0·35 0·08 – 0·10 – 0·10 0·01 – – – – 0·05 – – SM2 AOC45 4·5 800 0·38 0·33 0·08 – 0·00 – 0·09 0·01 – – – 0·04 – – – SM2 AOC47 4·5 850 0·41 0·30 – – – – 0·03 0·00 0·21 0·03 (DC) 0·01 – – – – 30SM3O D850-45 4·5 850 0·42 0·26 – – – – 0·09 0·01 0·16 0·02 (DC) <0·01 – – <0·01 – SM2-d AOC38 4·5 900 0·43 0·33 – – – – – <0·01 0·20 0·04 (DC) – – – – – SM2-d AOC34 4·5 1000 0·42 0·31 – – – – 0·00 0·00 0·23 0·04 (S+DC) – – – – – 1 SM, Starting material. 2 30SM3O: Mix of 30%SM3O+70%SM2. 3 Ca-features includes Ca-rich pools and skeletal crystals (DC: dendritic crystals, S: spherules). 4 Fe, Ti: Fe, Ti-rich oxides. 5 Phase proportions calculated with volatile melt composition obtained from GC-TCD analyses (see Table 5) * Phases were identified only. The capsules were then wrapped in a wet tissue to keep them cool and welded shut. The capsule weight was measured before and after the arc welding to detect leaks. For all experiments, the capsule was inserted in a ½ inch assembly made of a MgO-cylinder, a graphite furnace and an external NaCl sleeve. The temperature, accurate to ±5°C, was monitored using a type B thermocouple and a Eurotherm temperature controller. Experiments below 4 GPa were run for 1 week in a 200 t end-loaded piston cylinder at the RSES-ANU. Confining pressure was first applied to the main ram, then the piston was moved at ambient temperature up to ∼0·5 GPa to ensure good compaction of the experimental assembly. Afterwards, pressure was increased to the desired values and temperature was synchronously increased at a rate of 100°C min–1. Experiments at 4·5 GPa were conducted on a 500 t Harwood Engineering press. Pressure and temperature increase were automated and followed the same procedure as before. Piston travel was monitored throughout the duration of the experiments to detect any leak and friction decay (Hermann et al., 2016). All experiments were quenched by shutting off the power to the heater. Mineral identification and major element analyses Phase relations on the mounted and exposed run products were investigated with a scanning Zeiss UltraPlus analytical electron microscope fitted with a field emission gun (FE-SEM) at the Centre for Advanced Microscopy at the ANU. Phase compositions were determined using a JEOL 6400 SEM with energy-dispersive spectrometry (EDS) operating at 15 kV and 1 nA at the Centre for Advanced Microscopy at the ANU. Representative mineral and melt compositions are given in Supplementary DataTable S1 (supplementary data are available for downloading at http://www.petrology.oxfordjournals.org). Quantitative X-ray mapping was performed by wavelength-dispersive spectrometry (WDS) on a JEOL 8530 F electron microprobe at CMCA-UWA with an acceleration voltage of 15 kV and a beam current of 20 nA. The size of the images is 53 × 400 pixels, with a pixel step of 0·3 µm and a dwell time of 100 ms. Mass-balance calculations For each experiment, phase identification and mineral composition analyses were conducted over the whole capsule. Fe- and Mg-rich minerals are typically zoned, but the other elements show reasonably good consistency. For mass-balance calculations, the averaged phase composition was calculated by using analyses performed over the entire capsule. For zoned minerals such as garnet, the average of mineral compositions has been used in the calculation. The mass-balance calculation has been done using the solver function in Excel, by minimizing the difference between the known starting material composition and the recalculated bulk chemical composition from the mineral compositions. GAS ANALYSIS BY GAS CHROMATOGRAPHY Design of GC-TCD for experimental capsule analyses The retrieved gold capsules after the experiments were pierced and the gas was analysed using a GC-TCD. This separates the different components of a gas mix on the basis of their affinity with the stationary phase filling the column and as a function of the parameters chosen for the analysis (Table 3). The instrument is composed of a line and a 0·5 l volume that are connected to a sample holder, a vein-type vacuum pump (VARIAN DS 102), two MKS Baratron® absolute pressure gauges for P > 1 mbar (type 122B) and P < 2 mbar (type 170), a high-purity He cylinder and an Agilent Technologies 6850 gas chromatograph (GC) (Fig. 2a). The sample holder can accommodate either a small gas bottle for standards or experimental charges. The GC is equipped with a capillary column (30 m in length, 0·530 mm internal diameter) filled with HP-PLOT/Q stationary phase and a thermal conductivity detector (TCD). Data acquisition and processing were performed with the Agilent Technologies Chemstation software. Table 3: Parameters of the GC-TCD method for water and CO2 analysis in the experimental samples and standardization Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Table 3: Parameters of the GC-TCD method for water and CO2 analysis in the experimental samples and standardization Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Carrier gas Helium Carrier flow rate (ml min–1) 4·5 Inlet temperature (°C) 200 Detector temperature (°C) 250 Reference flow (ml min–1) 20 Make-up flow (ml min–1) 7 Split ratio 5:1 Oven profile Initial temperature (°C) hold for 2 min 40 Ramp (°C min–1) 20 Final temperature (°C) hold for 2 min 150 Fig. 2. View largeDownload slide (a) Analytical setup of the gas chromatograph adapted for the analysis of CO2 and H2O in experimental capsules. (b) Chromatogram of the analysis of CO/Air, CH4, CO2 and water balanced with He by GC-TCD with the method parameters given in Table 3. (c) CO2 and H2O characterization of the experimental gas, considered as a proxy for the fluid phase released from the subducting altered oceanic crust. The He-blank is compared with experiment AOC46 (3 GPa, 750°C) and AOC43 (3·5 GPa, 750°C) responses from gas chromatography analysis. Gas has massively exsolved from the melt during the quench of the high-pressure experiment (AOC43) leading to overall larger CO2 and H2O peaks. The X(CO2) of the exsolved fluid is higher in the high-pressure experiment AOC43 compared with experiment AOC46. Fig. 2. View largeDownload slide (a) Analytical setup of the gas chromatograph adapted for the analysis of CO2 and H2O in experimental capsules. (b) Chromatogram of the analysis of CO/Air, CH4, CO2 and water balanced with He by GC-TCD with the method parameters given in Table 3. (c) CO2 and H2O characterization of the experimental gas, considered as a proxy for the fluid phase released from the subducting altered oceanic crust. The He-blank is compared with experiment AOC46 (3 GPa, 750°C) and AOC43 (3·5 GPa, 750°C) responses from gas chromatography analysis. Gas has massively exsolved from the melt during the quench of the high-pressure experiment (AOC43) leading to overall larger CO2 and H2O peaks. The X(CO2) of the exsolved fluid is higher in the high-pressure experiment AOC43 compared with experiment AOC46. Procedure of experimental gas analysis by gas chromatography A typical GC-TCD analysis of an experimental sample lasts ∼5 h. Beforehand, the line and volume were flushed several times with He and evacuated overnight. Vacuum pressures are below 0·0001 mbar. The session starts with the analysis of several blanks (at least four) performed with high-purity He to check for the intensity and reproducibility of the background as well as the presence of possible contaminants. Air and water contamination is systematically observed in the blanks and appears on the chromatograms as a narrow peak eluted at ∼2·2 min for air and a fronting and tailing peak at 6·3 min for water. However, the areas of these peaks are very low (10–25 × 25 µV s) compared with that of the samples (> 250–300 × 25 µV s) and show little variation from one blank to another. The experimental capsule is heated with a heat gun to ∼250°C for 1–2 min to ensure that all free volatiles in the capsule are present as a gas phase. This is especially important for water, which is liquid at room conditions and may remain in the porosity of the experimental charge. Gas from unheated capsules has indeed returned suspiciously high X(CO2). Carbonates from the GC experiments show euhedral shapes and even dendritic crystals were preserved after the heating. Large and small carbonate crystals are equally preserved. Heated experimental charge and unheated experiments performed at the same P–T conditions show the same textures. This demonstrates that no significant carbonate dissolution occurred in consequence of the ∼250°C heating of the capsules. Then the capsule is pierced with a needle under vacuum (Fig. 2). The gas pressure is the difference between the vacuum pressure at room temperature (∼25°C) and the pressure read after piercing and cooling back to room temperature. One and a half hours is needed for the gas pressure in the line to stabilize after heating. The volume and line are topped with He to reach ∼1440 mbar. The gas sample and He mixture is then heated to ensure its homogenization. After cooling to room temperature, the gas mixture is analysed by GC-TCD at least five times. The average area of each peak is calculated with a minimum of five analyses. The relative standard deviation on the average peak area ranges from 0·6 to 2·1% for water and from 0·8 to 1·7% for CO2, attesting to the good reproducibility of the analyses. The validation of the gas analysis by GC-TCD necessitates that all potential gas compounds (CO, CH4, CO2 and water) are eluted at different times and thus are satisfactorily resolved in the chromatogram. For that purpose, a gas mixture composed of 5% CO2, 5% CO, 5% N, 5% O, 4% CH4, 4% H, balanced with He has been analysed with the same method as the samples. The capillary column HP-PLOT/Q used in this study cannot separate air and CO. Repeated He-blank analysis shows that air contamination is very low, ensuring that CO would be identified if present during the analysis. The chromatogram (Fig. 2b) shows that CO, CH4, CO2 and H2O are eluted at 2·222, 2·415, 3·055 and 6·264 min, respectively, which demonstrates that all compounds are satisfactorily separated at the conditions of the analyses. Standardization has been conducted to calibrate the relationship between the peak area and the gas partial pressure in the mixture. Details on the standardization of the gas analyses are given in the Supplementary Material. RESULTS Phase relationships Re-buffered and variable Fe3+ content experiments Experiments with 20–75% of oxidized starting material SM3O were all performed at 750°C, 3 GPa. They show the same phases as observed in experiment AOC13, carried out with SM2 alone (Table 2). In these experiments, a melt coexists with garnet + omphacite + coesite + phengite + epidote + dolomite and rutile is the only Ti-rich phase (Fig. 3a). In contrast, the experiment AOC30, composed of SM3O alone, contains the main paragenesis omphacite + phengite + epidote + kyanite + garnet + coesite + magnesite. In this experiment, melt was not observed. Garnet is present in a low proportion (∼4%, Table 2), magnesite is the sole carbonate present and rutile coexists with an Ti–Fe2+–Fe3+-oxide belonging to the magnetite–ulvöspinel solid solution (Fig. 3b). The experiments AOC12 and Re4, buffered with Re–ReO2, show a mineral assemblage composed of garnet + omphacite + coesite + phengite + epidote + dolomite and rutile (Fig. 3c). Melt was not present in these experiments. Fig. 3. View largeDownload slide BSE images in fO2-controlled experimental run products. (a) Experiment AOC29 with 75% oxidized starting material at 3·0 GPa and 750°C. (b) Experiment AOC30 with oxidized starting material only at 3·0 GPa and 750°C. (c) Experiment AOC12 with Re–ReO2 at 3·0 GPa and 750°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite. Fig. 3. View largeDownload slide BSE images in fO2-controlled experimental run products. (a) Experiment AOC29 with 75% oxidized starting material at 3·0 GPa and 750°C. (b) Experiment AOC30 with oxidized starting material only at 3·0 GPa and 750°C. (c) Experiment AOC12 with Re–ReO2 at 3·0 GPa and 750°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite. Phase relationships of the altered oceanic crust at the experimental conditions Experiments were performed at 2·5 GPa (700–850°C), 3 GPa (700–850°C), 3·5 GPa (700–800°C) and 4·5 GPa (700–1000°C). Run products are all characterized by a typical eclogitic paragenesis composed of garnet, omphacite and rutile (Fig. 4a). The solidus of the altered oceanic crust is located between 700 and 750°C at P ≤ 3·5 GPa and between 800 and 850°C at 4·5 GPa. The formation of a hydrous siliceous melt during the experiments is demonstrated by the presence of a glass sprinkled with many vesicles (Figs 5 and 6). From about 800 to 850°C, the glass is heterogeneous and contains CaCO3 as dendrites, sometimes tens of microns long, fillings of small cavities, or round pools in the melt (Fig. 6). The abundance of these features increases with increasing temperature. At all investigated pressures, the solidus is characterized by a decrease in omphacite, phengite, kyanite (when present), quartz or coesite and epidote, whereas garnet proportion tends to increase (Fig. 7). At P ≤ 3·5 GPa, the solidus coincides with the end of kyanite and magnesite stability fields. Above the solidus, phengite and quartz or coesite are then the next solid phases to disappear. Epidote is present as a major phase only at P = 2·5–3 GPa, 700–750°C. Above the solidus, epidote is mainly an accessory phase and has not been observed at all in the 4·5 GPa experiments. The stability field of dolomite decreases with increasing pressure above the solidus from 700 to 800°C at 2·5–4·5 GPa, where dolomite has been observed only as an accessory mineral in one experiment (AOC47). At P ≤ 3·5 GPa, dolomite is consistently replaced by Mg-calcite above the solidus. At T > 800°C, the solid assemblages are free of water. At P = 4·5 GPa, T > 800°C, the solid assemblage is free of any volatile. Fig. 4. View largeDownload slide (a) Experimental phase relationships in altered oceanic crust. (b) Fluid and melt compositions in the experimental products (see text for explanation). Numbers in black are the mole fraction of CO2 of the quench fluid, obtained from GC-TCD analyses. Numbers in red are the molar per cent of CO3 in the melt estimated from the starting material composition and mineralogy observed in the experiments. Numbers in bold black give the weight percentage of CO2 in the fluid or melt compared with total CO2 in the experiment. Fig. 4. View largeDownload slide (a) Experimental phase relationships in altered oceanic crust. (b) Fluid and melt compositions in the experimental products (see text for explanation). Numbers in black are the mole fraction of CO2 of the quench fluid, obtained from GC-TCD analyses. Numbers in red are the molar per cent of CO3 in the melt estimated from the starting material composition and mineralogy observed in the experiments. Numbers in bold black give the weight percentage of CO2 in the fluid or melt compared with total CO2 in the experiment. Fig. 5. View largeDownload slide Backscattered electron images of experimental mineral assemblages. (a) Subsolidus experiment AOC10 at 2·5 GPa, 700°C. (b) Melt-bearing experiments at 2·5 GPa, 800°C (AOC6). The presence of bubbly glass pools reflecting the exsolution of fluid during the quench should be noted. (c, d) Experiments AOC41 and AOC46 at 3·0 GPa, 700°C and 3·0 GPa, 775°C, respectively. (e) Subsolidus experiment AOC11 at 3·5 GPa, 700°C. (f) Subsolidus experiment D700-45 at 4·5 GPa, 700°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite; Mg-cc, Mg-rich calcite; Lw, lawsonite. Fig. 5. View largeDownload slide Backscattered electron images of experimental mineral assemblages. (a) Subsolidus experiment AOC10 at 2·5 GPa, 700°C. (b) Melt-bearing experiments at 2·5 GPa, 800°C (AOC6). The presence of bubbly glass pools reflecting the exsolution of fluid during the quench should be noted. (c, d) Experiments AOC41 and AOC46 at 3·0 GPa, 700°C and 3·0 GPa, 775°C, respectively. (e) Subsolidus experiment AOC11 at 3·5 GPa, 700°C. (f) Subsolidus experiment D700-45 at 4·5 GPa, 700°C. Ep, epidote; Ph, phengite; C, coesite; Do, dolomite; Grt, garnet; Om, omphacite; Mg, magnesite; Mg-cc, Mg-rich calcite; Lw, lawsonite. Fig. 6. View largeDownload slide BSE images of the Ca-rich features (indicated with the white arrow) observed in association with the glass. (a) Experiment AOC7 at 2·5 GPa, 850°C. (b) Experiment AOC21 at 3·0 GPa, 850°C. (c) Experiment AOC47 at 4·5 GPa, 850°C. (d–f) Mg, K and P X-ray quantitative maps (in element wt %). (g) Experiment AOC38 at 4·5 GPa, 900°C. (h) Experiment AOC34 at 4·5 GPa, 1000°C. In all these experiments, the Si-rich glass is sprinkled with bright, Ca-rich dendritic crystals (DC) or Ca-rich pools (Ca-P), reflecting the high solubility of carbonates in the melt at these conditions. Fig. 6. View largeDownload slide BSE images of the Ca-rich features (indicated with the white arrow) observed in association with the glass. (a) Experiment AOC7 at 2·5 GPa, 850°C. (b) Experiment AOC21 at 3·0 GPa, 850°C. (c) Experiment AOC47 at 4·5 GPa, 850°C. (d–f) Mg, K and P X-ray quantitative maps (in element wt %). (g) Experiment AOC38 at 4·5 GPa, 900°C. (h) Experiment AOC34 at 4·5 GPa, 1000°C. In all these experiments, the Si-rich glass is sprinkled with bright, Ca-rich dendritic crystals (DC) or Ca-rich pools (Ca-P), reflecting the high solubility of carbonates in the melt at these conditions. Fig. 7. View largeDownload slide The systematic change in mineral proportions at (a) 2·5 GPa, (b) 3·0 GPa, (c) 3·5 GPa and (d) 4·5 GPa. The different grey-scale of the markers represents the different starting materials (black, SM2; intermediate grey, 70SM3 + 30SM3O; light grey outline, SM2d). Fig. 7. View largeDownload slide The systematic change in mineral proportions at (a) 2·5 GPa, (b) 3·0 GPa, (c) 3·5 GPa and (d) 4·5 GPa. The different grey-scale of the markers represents the different starting materials (black, SM2; intermediate grey, 70SM3 + 30SM3O; light grey outline, SM2d). Only few experiments were performed at subsolidus conditions. Lawsonite in significant abundance (10%, Table 2), coexisting with magnesite, is restricted to P = 4·5 GPa and T = 700°C. Mineral compositions In all the experiments, most Fe–Mg-bearing phases (omphacite, magnesite, dolomite) and typically garnet show an important zoning in Fe and Mg (Figs 5 and 6d). As a result, the almandine component (XAlmandine) in garnet from one single experiment can vary from 0·65 to 0·40, making it difficult to interpret the Fe–Mg composition of minerals. As analyses on the rims of the grains usually return unreliable results, the XAlmandine in garnet given in this study should therefore be used with caution and considered as a maximum value for the equilibrium composition. Representative mineral compositions are given in the Supplementary Data (Tables S1–S6) and are reported in Fig. 8. Fig. 8. View largeDownload slide Evolution of mineral compositions with P–T conditions in the run products for (a) garnet, (b) omphacite and (c) celadonite. Fig. 8. View largeDownload slide Evolution of mineral compositions with P–T conditions in the run products for (a) garnet, (b) omphacite and (c) celadonite. Garnet Garnet from all experiments except AOC30 does not show significant Fe3+ content. XGrossular shows a general increase with pressure, especially at low temperature: at 700°C, XGrossular varies from 0·27 to 0·39 at 2·5 and 4·5 GPa, respectively. In the SM3O-bearing experiments, garnet has the same grossular content (XGrossular = 0·29–0·31; Fig. 8a) as the SM2-only experiment (AOC13) performed at the same P–T conditions. Garnet in the SM3O-only experiment (AOC30) shows a significantly lower grossular content (XGrossular = 0·19) and contains Fe3+ (XAndradite = 0·10). These garnets are also not zoned in Fe–Mg and do not show any inclusions in their cores. Clinopyroxene In most experiments, clinopyroxene is omphacite with 0·2 ≤ Na/(Na + Ca) ≤ 0·8 and no Fe3+ (Fig. 8) [following the classification of Morimoto (1988)]. AOC30, which contains only the oxidized starting material (100SM3O), is the only experiment showing clinopyroxene with a small component of aegirine (1–7%, 5% on average). The proportion of jadeite in omphacite increases with increasing pressure from 0·22–0·27 at 2·5–3·5 GPa to 0·37 at 4·5 GPa, 800°C. At fixed pressure, the proportion of jadeite in omphacite generally decreases with increasing temperature from 0·27 to 0·21 at 2·5 GPa and 0·37 to 0·33 at 4·5 GPa. In the SM3O-rich and Re-buffered experiments, omphacite crystals show similar jadeite contents (0·26–0·29) to those in the SM2-only experiment AOC13 (0·28). Carbonate phases Mg-calcite is CaCO3-rich (XCa = 0·76–0·82) with minor MgCO3 (XMg = 0·13–0·18) and FeCO3 components (XFe = 0·03–0·05). Dolomite and magnesite are highly zoned in Fe and Mg. Dolomite has an FeCO3 component ranging from XFe 0·15 to 0·31. Magnesite generally has an FeCO3 component ranging from 0·20 to 0·40, and a minor CaCO3 component below 0·06. Dolomite has similar composition in the SM3O-bearing and SM2-only experiments. Phengite The proportion of celadonite in phengite increases with pressure at fixed temperature from 0·4 at 2·5 GPa to 0·6 at 4·5 GPa, all at 700°C. At fixed pressure, the Ti content in phengite tends to increase with increasing temperature, and at fixed temperature, the Ti content in phengite decreases with increasing pressure. Epidote Epidote usually forms needle-shaped crystals that made analyses difficult. In experiments where chemical analyses were obtained, epidote shows a high clinozoisite component (0·67–0·77). At 3·5 GPa, 775°C, the clinozoisite component of the epidote reaches ∼0·92, which suggests that zoisite might actually the stable phase at these conditions. The highest Fe contents in epidote are observed in the Fe3+-rich experiments AOC30 and AOC12, with a clinozoisite component ranging between 0·42 and 0·53, respectively. Fe–Ti-oxides The Fe–Ti oxides observed in experiment AOC30 are composed of Fe2TiO2 (∼0·48) and Fe3O4 (∼0·52), given in molar proportions. Glass and fluid compositions Si-rich glasses Glasses are generally rich in vesicles and the electron microprobe analyses return totals around 80 wt % (Table 4), which attest to their volatile content or porosity. Si-rich glasses have trondhjemite–granodiorite compositions at 2·5 GPa and then become enriched in normative orthoclase with increasing pressure [classification from Barker (1979); Fig. 9]. The SiO2 content of the glasses ranges from 72 to 81 wt % (analyses normalized to 100 wt% without the volatiles H2O and CO2) and Al2O3 from 11 to 17 wt %. The K2O content of the melt ranges from 2·4 to 6·2 wt %. Table 4: Average glass compositions in experimental runs Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 * Melt composition recalculated with CO2 and H2O content measured by GC-TCD. n.d.: not determined. Table 4: Average glass compositions in experimental runs Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 Name P (GPa) T (°C) Glass n SiO2 Al2O3 TiO2 MgO FeO CaO Na2O K2O P2O5 CO2 H2O Total AOC5 3·5 800 Si-glass 6 62·9 11·4 0·3 0·1 0·4 1·4 2·5 3·0 n.d. 81·8 AOC6 2·5 800 Si-glass 7 59·0 13·8 0·2 0·4 0·6 2·0 1·8 2·2 n.d. 80·0 AOC7 2·5 850 Si-glass 5 58·1 13·9 0·4 0·2 0·3 1·7 3·4 2·2 n.d. 80·1 AOC7 2·5 850 C-features 2 25·2 6·4 0·3 3·8 3·9 25·7 2·4 1·2 n.d. 68·9 AOC8 3·5 750 Si-glass 8 65·0 10·7 0·1 0·0 0·4 1·4 1·8 2·2 n.d. 81·7 AOC9 2·5 750 Si-glass 5 60·7 12·5 0·1 0·2 0·5 1·4 3·5 2·1 n.d. 81·0 AOC13 3 750 Si-glass 4 63·2 11·0 0·2 0·1 0·2 1·4 2·4 2·1 n.d. 80·7 AOC46 3 775 Si-glass 5 65·6 12·6 0·3 0·2 0·6 1·5 1·5 1·9 n.d. 84·3 AOC46* 55·8 10·7 0·2 0·2 0·5 1·3 1·3 1·6 n.d. 1·4 27·0 100·0 AOC43 3·5 775 Si-glass 4 61·8 9·5 0·2 0·1 0·3 0·9 1·8 2·5 n.d. 77·0 AOC43* 44·1 6·8 0·2 0·0 0·2 0·6 1·3 1·8 n.d. 5·0 40·0 100·0 AOC15 3 800 Si-glass 6 58·7 11·8 0·3 0·3 0·5 1·1 2·0 2·6 n.d. 77·0 AOC15 3 800 C-features 2 15·0 6·3 0·7 2·8 4·7 32·1 2·3 0·6 n.d. 64·6 AOC21 3 850 Si-glass 5 61·7 13·3 0·6 0·4 0·6 2·1 1·1 2·0 0·2 81·7 AOC34 4·5 1000 Si-glass 3 63·6 10·9 1·1 0·4 0·7 1·4 0·8 3·5 0·1 82·4 AOC34 4·5 1000 C-features 2 10·2 3·4 1·6 2·9 7·3 30·3 0·6 0·7 1·4 56·9 AOC38 4·5 900 Si-glass 4 68·3 9·2 0·7 0·0 0·4 0·9 0·9 4·6 0·0 85·0 AOC38 4·5 900 C-features 2 24·4 3·8 0·4 1·6 3·5 28·0 0·6 1·8 1·5 64·1 AOC47 4·5 850 Si-glass 8 57·3 9·5 0·5 0·1 0·3 0·8 1·5 4·7 n.d. 74·6 AOC47 4·5 850 C-features 2 8·8 1·7 0·5 5·6 5·0 30·3 0·9 1·1 n.d. 53·9 * Melt composition recalculated with CO2 and H2O content measured by GC-TCD. n.d.: not determined. Fig. 9. View largeDownload slide Melt compositions from this study and previous studies in the granite classification of Barker (1979). Fig. 9. View largeDownload slide Melt compositions from this study and previous studies in the granite classification of Barker (1979). Ca-rich pools and dendritic crystals The Ca-rich features are always associated with glass and are very small in size (except at 4·5 GPa, 850°C, few tens of microns; Fig. 6). Their chemical composition, given in Table 4, is usually contaminated by the surrounding Si-melt. However, they show high contents of FeO (3·9–11 wt %) and MgO (1·6–5·6 wt %; Fig. 6d). Their contents of P2O5 (1–2·4 wt %; Fig. 7f), TiO2 (0·3–1·65 wt %) and BaO (detected by EDS analyses) are also higher than what is observed in the silicate melt. They are characterized by very low totals (50–70 wt %), suggesting a high content in volatiles, presumably CO2. Gas In all the experiments analysed by GC-TCD, except Re4, the chromatograms show two peaks at 3·1 s and 6·3 s, which correspond to the elution time of CO2 and H2O, respectively, and a small peak at 2·2 s, which corresponds to air–N2 contamination (area below the peak is <5) (Fig. 2c). The total pressure of gas enclosed in the capsule is typically 7·4–11·6 mbar for 68–94 mg of starting material (Table 5). The mole fraction of CO2 is referred to as X(CO2). GC-TCD analyses of the aqueous fluid present at 700°C show a slight decrease of X(CO2) from 0·059 ± 0·003 to 0·038 ± 0·003 with increasing pressure from 3·0 to 3·5 GPa, respectively (Table 5, Fig. 4b). At 4·5 GPa, 700°C, the capsule leaked during quenching and thus the gas could not be analysed by GC-TCD. Above the solidus at 775°C, the gas present in the capsule is characterized by an X(CO2) of 0·035 ± 0·002 and 0·057 ± 0·003 at 3·0 and 3·5 GPa, respectively (Table 5, Fig. 4b). At 4·5 GPa, 850°C, the gas present in the capsule is characterized by an X(CO2) of 0·043 ± 0·001. In experiment Re4 (3·0 GPa, 750°C), the total pressure of gas in the capsule was 14·4 mbar for 74·3 mg starting material. The chromatogram shows three peaks at 2·2 s, 3·1 s and 6·3 s, which correspond to the elution time of air–N2, CO2 and H2O, respectively. The areas under the CO2 and H2O peaks yield an X(CO2) of ∼0·146 ± 0·006 (Table 5). Table 5: Gas pressure and composition in the experimental runs from GC analyses Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. 1 Pressure of gas in the capsule after piercing under vacuum. 2 Average of area under the peak for each gas. 3 The gas pressure is obtained from the calibration lines established from the standards (see Supplementary Material). 4 XCO2 = PCO2/(PCO2 + PH2O). 5 The mass of CO2 in the capsule is calculated from the ideal gas law. 6 Mass of water in the starting material. n.a., not applicable. Table 5: Gas pressure and composition in the experimental runs from GC analyses Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. Experiment: AOC41 2σ AOC40 2σ AOC43 2σ AOC42 2σ AOC46 2σ Re4 2σ Sample pressure1 11·45 0·20 11·63 0·20 11·61 0·20 7·41 0·20 9·11 0·20 14·40 0·20 n 6 7 5 5 6 5 Average area CO22 42·2 1·5 26·9 0·5 40·1 1·1 15·2 0·1 18·6 0·3 114·4 1·9 Average area H2O 317·7 8·7 327·4 6·4 313·9 6·7 205·4 3·8 248·8 2·5 330·3 11·3 PCO23 0·70 0·03 0·45 0·03 0·67 0·02 0·26 0·05 0·31 0·02 1·93 0·02 PH2O 11·11 0·44 11·48 0·02 11·01 0·02 7·18 0·03 8·64 0·01 11·3 0·2 XCO24 0·059 0·003 0·038 0·003 0·057 0·003 0·035 0·004 0·035 0·002 0·146 0·006 CO2 mass (mg)5 0·62 0·40 0·59 0·23 0·28 n.a. H2O mass (mg) 4·04 4·17 4·00 2·61 3·14 n.a. SM H2O mass6 4·64 4·23 5·27 5·29 5·85 n.a. 1 Pressure of gas in the capsule after piercing under vacuum. 2 Average of area under the peak for each gas. 3 The gas pressure is obtained from the calibration lines established from the standards (see Supplementary Material). 4 XCO2 = PCO2/(PCO2 + PH2O). 5 The mass of CO2 in the capsule is calculated from the ideal gas law. 6 Mass of water in the starting material. n.a., not applicable. DISCUSSION Approach to equilibrium All experiments contained a fluid phase, facilitating well-crystallized minerals with euhedral shapes. There is no evidence for meta-stable phases and no variation of mineral assemblages throughout the capsule. Phase and melting relations are in agreement with previous studies (see below). At a given pressure, mineral and melt proportions evolve consistently with increasing temperature (Fig. 7), which suggests that the experimental assemblages approached equilibrium. Because of the overall low temperature of the experiments, there is some variation in mineral compositions. Garnet cores are enriched in almandine compared with the rims, suggesting that garnet nucleates early in the experiments, probably eased by the presence of Fe as Fe2SiO4. Nevertheless, garnet rim compositions show a consistent increase in grossular content with increasing pressure from 2·5 GPa to 4·5 GPa (Fig. 8), showing that garnet equilibrated during the experiments. The partition coefficient of Fe and Mg between clinopyroxene and garnet rims overall decreases with increasing temperature at a given pressure and starting material composition. The Si and Ti contents in phengite follow the trend with pressure and temperature described previously by Hermann & Spandler (2008) and Auzanneau et al. (2009). Carbonate stability At subsolidus temperatures, dolomite coexists with magnesite at P = 2·5–3·5 GPa. At 4·5 GPa, magnesite alone is observed with lawsonite at 700°C and then is replaced by dolomite at 800°C (Fig. 10). These parageneses agree with natural assemblages observed in HP and UHP rocks where magnesite and dolomite are typically described in equilibrium with eclogitic assemblages (Compagnoni & Rolfo, 2003; Smit et al., 2008; Zhang et al., 2009). The coexistence of lawsonite and magnesite is also in agreement with thermodynamic modelling on carbonated basalt, where these two phases coexist between 3 and 5 GPa at low temperature (Kerrick & Connolly, 2001). This suggests that our experimental setup and the fO2 conditions imposed by the experimental assembly (close to Ni–NiO equilibrium) yield results that are consistent with natural and modelled observations. By comparison with experiments in the CaCO3–CaMg(CO3)2 system, this study shows that the addition of Fe2+ above 3·5 GPa does not lead to a significant shift of the reaction dolomite + aragonite = Mg-calcite, obtained in the Fe-free system (Hermann et al., 2016). Our experimental results differ from previous work by Poli et al. (2009), in which graphite was found as the main C-bearing phase in the oceanic crust at the same conditions. However, the breakdown of oxalic acid, used in the experiments of Poli et al. (2009) to fix the C/H ratio of the fluid, can involve an excess of H2 or even the crystallization of graphite (Poli et al., 2009; McCubbin et al., 2014), both potentially driving the experiments to lower fO2 conditions. As noted by Poli et al. (2009), carbonate stability and fluid composition are closely linked to oxygen fugacity. Fig. 10. View largeDownload slide Experimental phase relationships for carbonate phases from this study and previous studies. Fig. 10. View largeDownload slide Experimental phase relationships for carbonate phases from this study and previous studies. Controls on oxygen fugacity The experiments AOC 28, 25 and 29 were composed of different proportions of starting materials SM2 and SM3O with different amounts of ferric iron. They all show the same phase relations as the experiment conducted with ferrous iron only (SM2; AOC13) (Table 2). Phase proportions in AOC29 (75%SM3O) are slightly different from those in AOC28 (20%SM3O) and AOC25 (50%SM3O), with lower modes of melt and garnet (Table 2). This suggests that the reducing capacity of the assembly with a graphite heater is very high and is able to counterbalance the addition of Fe3+ up to significant amounts of Fe3+ content (here up to 50–75 wt % of oxidized starting material SM3O). Small amounts of Fe3+ present in the starting material can also be buffered by the major minerals of the HP assemblage such as epidote, omphacite and garnet, without changing the phase relationships. On the other hand, no graphite was found in our experiments, suggesting that oxygen fugacity was not sufficiently low to reduce CO2 to graphite. Also, no Fe loss to the gold capsule was observed, indicating that no significant reduction of FeO occurred. From these observations we conclude that the intrinsic oxygen fugacity in the applied assembly is close to NNO as suggested previously (Hermann & Spandler, 2008). The experiment AOC30, composed of SM3O alone, shows drastically different phase relationships, with magnesite as the only carbonate phase, no melt and Fe–Ti-rich oxides with a high Fe3+ content (Fig. 3b). Garnet is present only as a minor phase (Supplementary Data) and garnet and clinopyroxene display small amounts of andradite and acmite components. Epidote is slightly enriched in Fe compared with experiment AOC13. These observations suggest that the high fO2 imposed by the starting material could no longer be buffered by the experimental assembly or mineral assemblage, but was accommodated by the crystallization of a different mineral paragenesis. Garnet in AOC30 is not zoned in Fe–Mg and does not contain inclusions, which suggests that high fO2 delays garnet nucleation, compared with low oxygen fugacity experiments where abundant fayalite seeds are available. The absence of melt is interpreted to result from the stability of C as CO2 gas in the fluid at high O-fugacity conditions, thus decreasing the activity of water and shifting the solidus to higher temperatures. The higher CO2 content of aqueous fluids at high-fO2 conditions is in line with the estimates of Molina & Poli (2000) comparing fluid composition at Hematite–Magnetite (HM) and NNO buffering conditions in the oceanic crust. O-buffered experiments by the addition of a Re–ReO2 mix show an absence of melt and a low abundance of garnet compared with the other experiments at the same P–T conditions (Fig. 3c;Table 2). In term of gas composition, the pressure of gas enclosed in the capsule is higher (∼14 mbar) compared with non-buffered experiments with higher amounts of starting material (<11·6 mg) (Table 5). The gas is composed of CO2, H2O and an additional gas eluted at 2·2 s. The X(CO2) of the gas is very high (∼0·15) compared with the experiment without Re–ReO2 and performed at 3·0 GPa and 775°C (0·036–0·059) (Table 5). The high gas pressure as well as the presence of an additional gas in the chromatogram compared with Re–ReO2-free experiments suggests that Re combines with OH or C to become volatile at the P–T conditions of the experiments. This shows that Re-compounds are volatile or soluble in water or CO2-rich solutions at high pressures, which is in line with previous observations at P < 0·5 GPa (Chellappa et al., 2009). These experiments confirm that Re can be released in the fluid phase at high pressures in a subduction-zone context (Becker, 2000; Lassiter, 2003). Therefore, the Re–ReO2 buffer is not suitable to buffer fO2 in H2O- or CO2-bearing experiments as it can affect the gas activities and phase relations during the run. The absence of melt in Re4 and AOC12, consistent with previous observations, is interpreted to be the result of high-fO2 conditions produced by the Re–ReO2 buffer and the low water activity. Phase and melting relationships in the altered oceanic crust: comparison with previous experiments and thermodynamic modelling Phase relations in K2O–H2O–CO2-bearing altered oceanic crust The temperature limit of the lawsonite stability field is between 700–800°C at 4·5 GPa and <700°C at 3·5 GPa (Fig. 4a). Although fairly large, this range is similar to that found experimentally or by thermodynamic modelling of basaltic compositions (Poli & Schmidt, 1995; Schmidt & Poli, 1998; Kerrick & Connolly, 2001; Poli et al., 2009). The evolution of modal proportions at 4·5 GPa (Fig. 7d) suggests that garnet grows at the expense of lawsonite, following the de-carbonation reaction lawsonite+4 magnesite+SiO2=pyrope+dolomite+2 CO2+2 H2O. (1) Amphibole has not been observed in this study, which is consistent with the previously determined maximum pressure (∼2·5 GPa) for the stability of this mineral in basaltic systems (Pawley & Holloway, 1993; Poli & Schmidt, 1995). In contrast, epidote is widespread throughout the P–T range investigated in this study, either as a major or minor phase at 2·5–3·5 GPa and T ≤ 750°C or as an accessory phase mostly above the solidus (Fig. 4a). The stability of epidote-group minerals at high-pressure conditions might be enhanced by the presence of trace elements in the starting material. Phengite is the stable K-phase in all subsolidus experiments in agreement with previous studies (Schmidt et al., 2004). Melting and dehydration reactions The presence of melt at experimental conditions is assumed from the presence of a glass in the experimental products. The solidus temperature can be bracketed between 700 and 750°C at 2·5–3·5 GPa (Fig. 4a). The appearance of glass is marked by a decrease of phengite, omphacite and kyanite modes and slight increase in garnet mode with increasing temperature (Fig. 7a–c). Dolomite is the main carbonate phase present below and above the solidus and its mode remains stable throughout the melting process. In contrast, magnesite, present in very minor amounts at subsolidus conditions (<0·03; Table 2), does not survive the onset of melting. It is thus possible that carbonate phases also take part in the melting reaction at 2·5–3·5 GPa. These observations suggest that the melting reaction involves the consumption of phengite, omphacite and kyanite at 3·0–3·5 GPa, and the production of melt and garnet: omphacite+phengite+fluid+coesite±kyanite+carbonate=melt±garnet. (2) The first phases to disappear are kyanite and fluid, as demonstrated by the persistence of coesite, phengite and epidote above the solidus at these pressures. This suggests that there is no excess aqueous fluid phase above the solidus (i.e. the melt is water undersaturated). This is in agreement with textural observations that there are no vesicles with ‘fish egg’ textures as observed in other studies at the same conditions where fluid-saturated melts were present (e.g. Skora et al., 2015). At 4·5 GPa, phengite is strongly enriched in the celadonite component and omphacite in the jadeite component, stabilizing these phases to higher pressure (Kessel et al., 2005) and thus shifting the solidus to higher temperatures. The absence of phengite above the solidus at high pressure is consistent with previous experiments in K2O, H2O-bearing basaltic system showing the coexistence of melt and phengite at 740°C and 3·2 GPa and the absence of phengite at 850°C and 4 GPa (Schmidt et al., 2004). The absence of phengite above the solidus also results in the enrichment of the melt in normative orthoclase, shifting the melt composition into the K-rich granite domain compared with lower pressure experiments from this and previous studies (Fig. 9). Garnet mode above the solidus is relatively constant compared with lower pressures (Fig. 7). Carbonate phases are observed only as accessory minerals at 850°C and disappear completely at higher temperature. This suggests that the melting–dehydration reaction now clearly involves a solid carbonate phase and is omphacite+phengite+dolomite+fluid=melt±garnet. (3) The previous, experimental estimations for the solidus in various chemical systems including either H2O, CO2 or K2O are reported in Fig. 11. The range of solidus temperatures below 3·5 GPa is similar to those observed in K2O, H2O-bearing basaltic compositions (Schmidt et al., 2004; Carter et al., 2015), as well as CO2-bearing, K2O-free basaltic compositions (solidus located between 700 and 750°C between 2·0 and 3·5 GPa; Yaxley & Green, 1994). However, the solidus temperatures found in this study at P ≥ 3·5 GPa are ∼50°C below experimental studies in K2O, H2O-bearing (Schmidt et al., 2004) and K2O-free, H2O-bearing basaltic compositions (Kessel et al., 2005). This might be explained by the lower water content in the first study (1·8 wt % compared with 6–7 wt % in this study) and the absence of K2O and phengite in the second study (Kessel et al., 2005). The solidus temperature determined here at 3 GPa occurs at a slightly lower temperature (700–750°C) than in the experiments of Skora et al. (2015) on low-carbonate pelite (LC) containing ∼7 wt % water (760–800°C) and those performed by Carter et al. (2015) (∼750°C, 3 GPa) on an altered oceanic crust composition (K2O = 0·14 wt %). For Skora et al. (2015), this difference in the solidus temperature is well explained by the higher Fe3+ content of their starting material (Fe3+/Fetot = 0·46), combined with a different experimental assembly buffering at slightly higher fO2 (above NNO). The experiments performed at different fO2 shown in this study indeed demonstrate that fO2 is critical in controlling the onset of the melting reaction, and experiments with the highly oxidized starting material (AOC30) or the experiment buffered at high fO2 (AOC12) show a solidus shifted to higher temperature conditions. The slight difference in melting temperature compared with the experiments of Carter et al. (2015) can again be assigned to differences in fO2 conditions, as demonstrated by the occurrence of Fe–Ti oxides and the difficulty of nucleating garnet in their experiments. Our results differ from the previous experiments on H2O–CO2–basaltic systems by Poli et al. (2009), where only one experiment, performed at fO2 conditions of NNO, showed a glass at 2·2 GPa, 730°C. The low solidus temperature found in our study, which is consistent with CO2-free and CO2-bearing systems at least at P < 4 GPa (Yaxley & Green, 1994; Poli & Schmidt, 1995; Schmidt et al., 2004), shows that the presence of CO2, when trapped in carbonate phases below the solidus, does not affect significantly the water activity in the system. This contrasts with high-fO2 conditions, which lead to dissolution of carbonates and fluids with higher amounts of CO2, and thus an increase in the solidus temperature. Fig. 11. View largeDownload slide Comparison of solidus temperatures for mid-ocean ridge basalt, altered oceanic crust and pelite systems. For intermediate to hot subduction geotherms (Syracuse et al., 2010) the subducted altered oceanic crust is able to undergo fluid fluxed partial melting at sub-arc depths (3–4·5 GPa). P–T paths of the slab surface beneath the Lesser Antilles (a) and Sunda (b) arc segments are from Syracuse et al. (2010). Fig. 11. View largeDownload slide Comparison of solidus temperatures for mid-ocean ridge basalt, altered oceanic crust and pelite systems. For intermediate to hot subduction geotherms (Syracuse et al., 2010) the subducted altered oceanic crust is able to undergo fluid fluxed partial melting at sub-arc depths (3–4·5 GPa). P–T paths of the slab surface beneath the Lesser Antilles (a) and Sunda (b) arc segments are from Syracuse et al. (2010). Nature and H2O–CO2 composition of the fluid phase at experimental conditions Aqueous fluid At 700°C (3–4·5 GPa), the volatile-rich phase is close to an aqueous fluid (scheme 1, Fig. 1), as demonstrated by the high water content of the gas escaping the capsule [X(CO2) = 0·038–0·059 at 3–3·5 GPa; Table 5] and the absence of glass in the experimental charges. The amount of fluid released determined by GC-TCD analyses indicates that about 70 wt % (at 3 GPa) and 80 wt % (at 3·5 GPa) of the carbon present in the starting material was retained in the residual assemblage (Fig. 4b). At 4·5 GPa, 700°C, the capsule leaked during quenching and thus the gas could not be analysed by GC-TCD. Nevertheless, SEM investigation of the charge, as well as mass-balance calculations, shows that a high amount of magnesite was present (∼5%; Table 2) and thus ∼90 wt % of the C was retained in the residue (Fig. 4b). The calculated X(CO2) is very low at 0·02–0·03, following the trend established at lower pressure. These results are in excellent agreement with calculated fluid compositions based on experiments that predict a decrease of X(CO2) in the fluid below ∼0·05 with increasing pressure above 2·5 GPa at NNO-buffered oxygen fugacity (Poli et al., 2009). Thermodynamic modelling of fluid compositions provided significantly lower X(CO2) ≤ 0·01 at sub-arc depths in an altered basalt composition (Connolly, 2005). The thermodynamic modelling was performed with a mixing model between molecular CO2 and H2O, which was consistent with experimental studies below 1·6 GPa on calcite solubility in water, showing that CO2aq was the main carbon-rich species (Caciagli & Manning, 2003). However, the higher solubility of carbonate in our experiments might suggest the presence of dissolved carbonate ions in the complex aqueous fluid at the higher pressures of our run conditions. This is supported by evidence of carbonate dissolution from fluid inclusion studies in natural UHP rocks (Frezzotti et al., 2011) and by recent thermodynamic modelling of C speciation in high-pressure fluids in equilibrium with eclogitic minerals (Sverjensky et al., 2014). To our knowledge, our data represent the first direct measurements of the CO2 concentration in aqueous subduction-zone fluids. Hydrous silicate melt The observed phase relations indicate that a single hydrous silicate melt (and no excess aqueous fluid) was present above the solidus. The volatile content of this melt can be constrained from the amount of observed H2O- and CO2-bearing phases (Table 2) and from the quantitative analysis of gas. The mass of volatiles coming out of the capsules, obtained from the GC-TCD analyses, can be used to estimate the proportion of gas exsolved during the quench. A minimum estimate is calculated assuming that there is no more CO2 in the glass. A maximum estimate considers that all the carbonates were dissolved in the melt. These calculations show that 60–80 wt % of gas was exsolved from the melt at 3·5 GPa and 40–58 wt % at 3·0 GPa. This observation demonstrates the importance of analysing the fluid composition by direct methods, here by GC-TCD, to well constrain the volatile budget of the experiments, as the glass composition may not represent the composition of the melt. At T > 700°C, the exsolved gas is characterized by an X(CO2) of 0·035 ± 0·002 and 0·057 ± 0·003 at 3·0 and 3·5 GPa, respectively (Table 5, Fig. 4b). It is more difficult to assess the CO2–H2O composition of the hydrous siliceous melt than of the fluids at subsolidus conditions because it is very difficult to measure the volatile composition of the small, bubble-rich glass pools by any analytical technique. The minimum X(CO2) in the hydrous siliceous melt can be estimated from the GC-TCD measurement and considering CO2Glass = 0, as X(CO2)Melt=CO2Gas/44H2OGlass/18+H2OGas/18+CO2Gas/44. H2OGas and CO2Gas are the mass of water and CO2 in the gas phase and are calculated from GC-TCD analysis by converting the gas partial pressure to mass using the ideal gas law. H2OGlass is obtained by mass balance using the proportion of hydrous siliceous melt and hydrous minerals in the experiments and the total mass of water in the starting material. The resulting minimum X(CO2) of the hydrous siliceous melt present at experimental conditions is thus >0·020 at 3·0 GPa and >0·045 at 3·5 GPa, 775°C (Fig. 4b). These values are lower than the bulk-rock X(CO2) of these experiments (0·16), which is consistent with the observation that residual carbonates are present in these runs (Table 2). These X(CO2) values translate to a proportion of CO2 released in the fluid phase (mass of CO2 in the fluid or melt divided by the mass of CO2 in the starting material) ranging from >10 to >25% (Fig. 4). Increased solubility of CO2 in silicate melts with pressure and implication for fO2 in the mantle wedge Some of our super-solidus experiments show Ca- and CO2-rich features associated with the silicate glasses. Their origin can be (1) a quench product from an immiscible carbonatite liquid (scheme 3, Fig. 1), (2) an unmixing of dissolved carbonate liquid on quenching of a single carbonate-rich silicate-melt (scheme 2, Fig. 1), or (3) quench material of solutes that derive from the unmixing of an aqueous fluid phase from a melt during quenching. The description of different quench textures was briefly reviewed by Poli (2015). Previous experimental work suggested that dendritic crystals tend to form as a quench product of a liquid (Wyllie & Tuttle, 1959). Quench carbonatite melts usually form large spherical pools with their own internal organization of dendritic crystals (Martin et al., 2013; Skora et al., 2015). In contrast, quench material after an exsolved fluid appears in spherical micro-vesicles sprinkled throughout the glass (Skora et al., 2015). At 850°C, 4·5 GPa, only accessory (<2%) residual carbonates and phengite (only in D850-45) were observed. The glass contains long Ca-rich dendritic crystals (Fig. 6) characterized by relatively high P, Ba and Ti contents. These dendritic crystals are interpreted to have formed from a carbonatite liquid exsolved from the hydrous silicate melt during quenching. In the higher temperature experiments at 4·5 GPa, dendritic crystals are replaced by micro-vesicles filled with Ca-rich material, suggesting that a gas, rather than a liquid, might have been exsolved during the quench. The textures observed in this study are very similar to those reported by Poli (2015) in experiments on hydrous carbonated gabbro at 4·2 GPa, 870–900°C. It remains unclear whether a single, hydrous silicate melt with a significant carbonatite component or two separate melts existed at the experimental conditions. In any case, all these observations reflect an enhancement of carbonate solubility in the hydrous siliceous melt. At 4·5 GPa, 850°C, the near total dissolution of all the carbonate and the large proportion of the dendritic crystals in the glass pools (10–15%) indicates that 75–100 wt % of C from the starting material was hosted in the hydrous siliceous melt, resulting in an estimated X(CO2) > 0·08. At 4·5 GPa, 900°C, all the carbonate added to the starting material is dissolved in the hydrous siliceous melt coexisting with garnet + omphacite + coesite. This experiment was performed with a lower water content [X(CO2) = 0·27] to avoid formation of an excess fluid phase. As no carbonate mineral is present in this experiment, the X(CO2) of the melt is 0·27 and 100% of the total CO2 present in the starting material occurs in the hydrous siliceous melt (Fig. 1). It is worth noting that the increase in X(CO2) of the hydrous siliceous melt from >0·08 to 0·27 at 4·5 GPa, 850–900°C reflects the composition of the starting material and, thus, the increased capacity of the hydrous siliceous melt to incorporate carbonate species. The increase of carbonate solubility with pressure and temperature in hydrous siliceous melts is thus dramatic at P > 4 GPa. The carbonate-rich melts produced in the slab at sub-arc pressures are also likely to invade the mantle wedge. The oxidizing power of CO2 in melts was demonstrated from experiments (Simakin et al., 2012) and studies of natural magmatic systems that evolved in response to interaction with carbonate-rich rocks (Wenzel et al., 2002; Simakin et al., 2012). In these studies, the increase in fO2 of the melt results in Fe oxidation and crystallization of Fe3+-rich clinopyroxene or Mg-rich olivine. This is consistent with the high Fe content of the C-rich features observed in our study, compared with the Si-rich glasses (Table 4). CO2-rich melts are thus potentially an efficient transfer medium for oxidized species and could increase the input of Fe and C in subduction zones. The oxidation state of the sub-arc mantle has long been a subject of debate (Parkinson & Arculus, 1999; Aeolus Lee et al., 2005) and has strong implications in terms of melting relations (Foley, 2011), redox budget (Evans, 2012; Evans et al., 2012) and the generation of Cu–Au deposits (e.g. Evans & Tomkins, 2011). Evidence for an oxidized sub-arc mantle includes primitive arc magma compositions and fO2 estimations from minerals forming arc basalts and associated peridotites (e.g. Parkinson & Arculus, 1999; Rowe et al., 2009; Evans et al., 2012). The oxidation of the sub-arc mantle is interpreted as the result of interaction with oxidized fluids coming from the subducting slab and usually considered as aqueous or silica-rich fluids (Evans, 2012). Our study suggests that CO2-rich melts formed at sub-arc depths in the subducting altered oceanic crust could also be responsible for the oxidation of the sub-arc mantle. IMPLICATIONS FOR CO2 AND H2O TRANSFER AT SUBDUCTION ZONES Our study suggests that the three critical factors controlling carbon recycling at subduction zones from altered oceanic crust are oxygen fugacity conditions, flushing of aqueous fluids from the dehydration of underlying units and the temperature on top of the slab at sub-arc pressures. The alteration of the oceanic crust evolves with depth from oxidizing to rather reducing conditions (Alt & Honnorez, 1984; Staudigel, 2003). Nevertheless, the mineral parageneses and compositions in blueschist and eclogite rocks generally suggest that high-pressure metamorphism, and associated devolatilization, then drives the fO2 towards reducing conditions. Rutile is the most common Ti-rich phase in high-pressure metabasites, whereas ilmenite or other Fe–Ti oxides are rarely observed unless at very low degrees of metamorphism (Galvez et al., 2012). In addition, garnet rarely shows a high andradite component. The study of Groppo & Castelli (2010) interpreted the Fe3+/ΣFe zoning in garnet as a decrease in fO2 across the FMQ buffer during prograde lawsonite destabilization in metabasites. Cao et al. (2011) estimated the fO2 of massive eclogite from the North Qilian suture zone (NW China) at ΔFMQ +2·5. The good consistency between the parageneses observed in our experiments and in natural rocks, as well as fO2 estimations in natural eclogites, suggests that the fO2 conditions of our experiments are relevant to subducted, altered oceanic crust at sub-arc conditions. Nevertheless, some occurrences of very oxidized rocks, associated with a peculiar mineralogy, are locally described in HP complexes; for example, the Mn ore of Praborna in the Italian Western Alps formed at 2 GPa, which records ΔFMQ up to 12·7 (Tumiati et al., 2015), and the epidote-talc bearing eclogites from the Sulu Terrane, which record fO2 conditions at ΔFMQ = 4·5 (Mattison et al., 2004). In such rare rock types, CO2 extraction owing to fluid–rock interaction is expected to be much more efficient. In the past, fluid-present melting of altered basalts has not been regarded as feasible because thermal models predicted temperatures too low for this to happen at sub-arc depths (Peacock, 1996). However, in the newest set of thermal models (Syracuse et al., 2010), a large number of subduction zones reach temperatures where the top of the altered basalt layer might undergo fluid-fluxed partial melting (Fig. 11). In addition, the wet solidus of altered basalt determined here is very similar to the wet solidus of sediments. Studies on K, Th and light rare earth element recycling in subducted sediments have demonstrated that partial melting is necessary in most subduction zones, suggesting that top slab temperatures can reach 700–900°C (Hermann & Spandler, 2008; Plank et al., 2009) and supporting partial melting in the altered oceanic crust as a likely process in subduction zones. Our experiments constrain water-fluxed melting of one specific type of altered oceanic crust, a K-enriched ‘supercomposite’ (Staudigel et al., 1989). The amount of carbonate and K-phases is highly variable depending on the depth and processes of metasomatism, as summarized in the section above on alteration of the oceanic crust. Although these highly variable rock compositions have not been investigated in this study, the obtained phase and melting relations shed some light on the devolatilization of these rocks. The solid phases at subsolidus conditions just before the onset of melting are garnet, clinopyroxene, coesite, phengite, carbonate ± kyanite, in agreement with a wide range of composition between sediments and basalts (Schmidt et al., 2004; Hermann & Spandler, 2008; Carter et al., 2015). Thus, different bulk compositions will result only in different proportions of these minerals (with variable compositions) but will not modify melting reactions (2) and (3). The extent of melting will then depend first on the amount of external aqueous fluid input and second on the order of exhaustion of the solid phases participating in the melting reactions. The melting temperature is expected to vary with phase composition, particularly the Mg# of the bulk-rock and the Na/Ca of the omphacite. In the absence of any external fluid input at sub-arc depths, the extent of melting of altered oceanic crust will be controlled by the amount of phengite and epidote, as they are the only stable hydrous phase at the solidus of these rocks. In this scenario the efficiency of carbon removal from oceanic crust will depend on the ratio between phengite + epidote and carbonate, and thus on the extent and type of alteration of the oceanic crust. The most recent estimations of carbon fluxes in subduction zones by Kelemen & Manning (2015) suggest that most carbon subducted along with serpentinite, sediments and the altered oceanic crust is likely to be brought back into the mantle wedge via fluids migrating from the slab. The main process proposed by the above researchers is the dissolution of carbonate minerals, calcite–aragonite, into the aqueous fluids percolating through the different lithologies of the slab. Our experimental results, combined with other experimental studies at similar conditions (Carter et al., 2015; Poli, 2015; Skora et al., 2015), demonstrate that partial melting of the subducting, crustal lithologies is possible at least along warm and hot geotherms (Fig. 11) and constitutes an alternative, effective way to mobilize carbon from the slab towards the mantle wedge. Furthermore, the model put forward by Kelemen & Manning (2015) necessitates diffusive transport of the percolating aqueous fluids throughout the slab lithologies, which is difficult to reconcile with natural observations showing that aqueous fluid flow is rather channelized than pervasive (e.g. Spandler et al., 2011). In terms of fluid transport, carbonatite melts are characterized by very low viscosity, especially at high-pressure conditions (e.g. Kono et al., 2014). The dissolution of carbonate in hydrous silicate melts would probably decrease their viscosity, thus helping in the migration of those fluids from the top of the slab to the mantle wedge. The time scales of deep carbon cycling critically depend on the main process of carbon release. Volatile recycling through arc magmatism takes less than 10 Myr (Morris et al., 1990). Residual carbonate in eclogite may result in melting at the Transition Zone or even at lower mantle conditions (Rohrbach & Schmidt, 2011) and recycling through intra-plate magmatism takes of the order of 100 Myr. The time scale for subduction to the lower mantle and recycling through plumes and mid-ocean ridge magmatism probably exceeds 1 Gyr (Dasgupta & Hirschmann, 2010). As altered oceanic crust is the main reservoir of subducted carbon, the efficient CO2 extraction by partial melting of subducted altered oceanic crust over short time scales of <10 Myr has important consequences for the long-term evolution of CO2 in the atmosphere (Berner et al., 1983). Interestingly, the range of X(CO2) of ∼0·02–0·08 mol % produced during wet melting at sub-arc depths (Fig. 4b) translates into a range of H/C mass ratios of ∼1–3. These values straddle the average H/C mass ratio of 1·92 of the exosphere (Hirschmann & Dasgupta, 2009). It needs to be evaluated, with future studies, whether wet melting of carbonated eclogite has been a dominant process throughout Earth’s history that could have fractionated C and H in the way that they are now present in the exosphere or whether this is just a coincidence. CONCLUSIONS (1) Solidus temperatures for the altered oceanic crust are between 700 and 750°C at 2·5–3·5 GPa and 800–850°C at 4·5 GPa, similar to those for CO2-free, K-bearing basaltic systems. This provides evidence for low solubility of CO2 in aqueous fluids in equilibrium with carbonate-bearing eclogite residues. (2) We have directly measured such aqueous fluids using gas chromatography, confirming a low X(CO2) of 0·059 ± 0·003 at 3 GPa, 700°C and 0·038 ± 0·003 at 3·5 GPa, 700°C. Although these values are low in absolute terms, they are still significantly higher than calculated values at the same conditions, which are generally below 0·01. This indicates that fluids are probably complex and might have also significant amounts of dissolved carbonate ions at sub-arc conditions. (3) Different proportions of oxidized versus reduced materials, as well as Re–ReO2, have been used to investigate the effect of fO2 on carbonate stability in the altered oceanic crust. It is shown that the reducing power of the assembly set the fO2 around NNO, even when up to 75% of oxidized material is used in the starting material. Experiments buffered with Re–Re2O suggest that Re gas compounds mix with CO2 and H2O, confirming that Re can become volatile at subduction conditions. An experiment carried out with oxidized material suggests that high fO2 stabilizes CO2 gas over carbonates at subsolidus conditions, thus increasing the solidus temperature at sub-arc depths and delaying the onset of partial melting to higher temperatures in subduction zones. (4) Partial melting of altered oceanic crust is possible in subduction zones along intermediate to hot geotherms for fO2 close to NNO. Carbonate solubility in these hydrous melts increases with depth, leading to efficient extraction of CO2 from the subducted slab at sub-arc conditions. Significant amounts of subducted carbon can be recycled back to the atmosphere via arc magmatism on relatively short time scales of less than 10 Myr. ACKNOWLEDGEMENTS The authors benefited from thoughtful and useful reviews by Ron Frost and Frieder Klein. We would like to thank Jim Beard for his editorial handling. James Tuff and Mike Shelley are acknowledged for early technical development on the gas chromatograph at RSES. We thank Greg Yaxley for comments, and Dave Clarke and Janet Hope for fruitful discussions and technical support for the gas chromatograph. Dave Clarke and Dean Scott are thanked for their technical support in the experimental laboratory. Malcolm Roberts and Bob Rapp are thanked for their assistance with the EPMA at CMCA-UWA and the SX100 at the RSES-ANU, respectively. FUNDING The authors acknowledge the facilities, and the scientific and technical assistance of the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy, Characterisation & Analysis, The University of Western Australia, a facility funded by the University, State and Commonwealth Governments. This work has been financially supported by the Australian Research Council. SUPPLEMENTARY DATA Supplementary data for this paper are available at Journal of Petrology online. REFERENCES Aeolus Lee C.-T. , Leeman W. P. , Canil D. , Li Z.-X. A. ( 2005 ). Similar V/Sc systematics in MORB and arc basalts: implications for the oxygen fugacities of their mantle source regions . Journal of Petrology 46 , 2313 – 2336 . Google Scholar CrossRef Search ADS Ague J. J. , Nicolescu S. ( 2014 ). Carbon dioxide released from subduction zones by fluid-mediated reactions . Nature Geosciences 7 , 355 – 360 . Google Scholar CrossRef Search ADS Akaishi M. , Yamaoka S. ( 2000 ). Crystallisation of diamond from C–O–H fluids under high-pressure and high-temperature conditions . Journal of Crystal Growth 209 , 999 – 1003 . Google Scholar CrossRef Search ADS Akaishi M. , Shaji Kumar M. D. , Kanda H. , Yamaoka S. ( 2000 ). Formation process of diamond from supercritical H2O–CO2 fluid under high pressure and high temperature conditions . Diamond and Related Materials 9 , 1945 – 1950 . Google Scholar CrossRef Search ADS Alt J. C. , Honnorez J. ( 1984 ). Alteration of the upper oceanic crust, DSDP site 417: mineralogy and chemistry . Contributions to Mineralogy and Petrology 87 , 149 – 169 . Google Scholar CrossRef Search ADS Alt J. C. , Teagle D. A. H. ( 1999 ). The uptake of carbon during alteration of ocean crust . Geochimica et Cosmochimica Acta 63 , 1527 – 1535 . Google Scholar CrossRef Search ADS Alt J. C. , Laverne C. , Vanko D. A. , Tartarotti P. , Teagle D. A. H. , Bach W. , Zuleger E. , Erzinger J. , Honnorez J. , Pezard P. A. ( 1996 ). Hydrothermal alteration of a section of upper oceanic crust in the eastern equatorial Pacific: a synthesis of results from site 504 (DSDP legs 69, 70, AND 83, and ODP legs 111, 137, 140, and 148). In: Alt, J. C., Kinoshita, H., Stokking, L. B. et al. (eds) Proceedings of the Ocean Drilling Program, Scientific Results , 148 . Ocean Drilling Program , pp. 417 – 434 . Alt J. C. , Schwarzenbach E. M. , Früh-Green G. L. , Shanks W. C. , Bernasconi S. M. , Garrido C. J. , Crispini L. , Gaggero L. , Padrón-Navarta J. A. , Marchesi C. ( 2013 ). The role of serpentinites in cycling of carbon and sulfur: seafloor serpentinization and subduction metamorphism . Lithos 178 , 40 – 54 . Google Scholar CrossRef Search ADS Arcay D. , Tric E. , Doin M. P. ( 2005 ). Numerical simulations of subduction zones: effect of slab dehydration on the mantle wedge dynamics . Physics of the Earth and Planetary Interiors 149 , 133 – 153 . Google Scholar CrossRef Search ADS Auzanneau E. , Schmidt M. W. , Vielzeuf D. , Connolly D. J. A. ( 2009 ). Titanium in phengite: a geobarometer for high temperature eclogites . Contributions to Mineralogy and Petrology 159 , 1 . Google Scholar CrossRef Search ADS Barker F. ( 1979 ). Trondhjemite: definition, environment, and hypotheses of origin. In: Barker F. (ed.) Trondhjemites, Dacites, and Related Rocks . Elsevier , pp. 1 – 12 . Becker H. ( 2000 ). Re–Os fractionation in eclogites and blueschists and the implications for recycling of oceanic crust into the mantle . Earth and Planetary Science Letters 177 , 287 – 300 . Google Scholar CrossRef Search ADS Becker K. , Sakai H. , Adamson A. C. , Alexandrovich J. , Alt J. C. , Anderson R. N. , Bideau D. , Gable R. , Herzig P. M. , Houghton S. , Ishizuka H. , Kawahata H. , Kinoshita H. , Langseth M. G. , Lovell M. A. , Malpas J. , Masuda H. , Merrill R. B. , Morin R. H. , Mottl M. J. , Pariso J. E. , Pezard P. , Phillips J. , Sparks J. , Uhlig S. ( 1989 ). Drilling deep into young oceanic crust, Hole 504B, Costa Rica Rift . Reviews of Geophysics 27 , 79 – 102 . Google Scholar CrossRef Search ADS Berner R. A. , Lasaga A. C. , Garrels R. M. ( 1983 ). The carbonate–silicate geochemical cycle and its effect on atmospheric carbon dioxide over the past 100 million years . American Journal of Science 283 , 641 – 683 . Google Scholar CrossRef Search ADS Cao Y. , Song S. G. , Niu Y. L. , Jung H. , Jin Z. M. ( 2011 ). Variation of mineral composition, fabric and oxygen fugacity from massive to foliated eclogites during exhumation of subducted ocean crust in the Norh Qilian suture zone, NW China . Journal of Metamorphic Geology 24 , 699 – 720 . Google Scholar CrossRef Search ADS Caciagli N. C. , Manning C. E. ( 2003 ). The solubility of calcite in water at 6–16 kbar and 500–800°C . Contributions to Mineralogy and Petrology 146 , 275 – 285 . Google Scholar CrossRef Search ADS Carter L. B. , Skora S. , Blundy J. D. , De Hoog J. C. M. , Elliott T. ( 2015 ). An experimental study of trace element fluxes from subducted oceanic crust . Journal of Petrology 56 , 1585 – 1606 . Google Scholar CrossRef Search ADS Chellappa R. S. , Somayazulu M. , Hemley R. J. ( 2009 ). Rhenium reactivity in H2O–O2 supercritical mixtures at high pressures . High Pressure Research 29 , 792 . Google Scholar CrossRef Search ADS Chepurov A. I. , Tomilenko A. A. , Zhimulev E. I. , Sonin V. M. , Chepurov A. A. , Kovyazin S. V. , Timina T. Y. , Surkov N. V. ( 2012 ). The conservation of an aqueous fluid in inclusions in minerals and their interstices at high pressures and temperatures during the decomposition of antigorite . Russian Geology and Geophysics 53 , 234 – 246 . Google Scholar CrossRef Search ADS Compagnoni R. , Rolfo F. ( 2003 ). Ultrahigh-pressure metamorphic rocks in the Western Alps. In: Carswell D. A. , Compagnoni R. (eds) Ultrahigh Pressure Metamorphism: EMU Notes in Mineralogy . Cambridge : Cambridge University press , pp. 206 – 243 . Connolly J. A. D. ( 2005 ). Computation of phase equilibria by linear programming: a tool for geodynamic modeling and its application to subduction zone decarbonation . Earth and Planetary Science Letters 236 , 524 – 541 . Google Scholar CrossRef Search ADS Dasgupta R. ( 2013 ). Ingassing, storage, and outgassing of terrestrial carbon through geologic time. In: Hazen R. M. , Jones A. P. , Baross J. A. (eds) Carbon in Earth. Mineralogical Society of America and Geochemical Society, Reviews in Mineralogy and Geochemistry 75 , 183 – 229 . Dasgupta R. , Hirschmann M. M. ( 2010 ). The deep carbon cycle and melting in the Earth’s interior . Earth and Planetary Science Letters 298 , 1 – 13 . Google Scholar CrossRef Search ADS Dasgupta R. , Hirschmann M. M. , Withers A. ( 2004 ). Deep global cycling of carbon constrained by the solidus of anhydrous, carbonated eclogite under upper mantle conditions . Earth and Planetary Science Letters 227 , 73 – 85 . Google Scholar CrossRef Search ADS de Leeuw G. A. M. , Hilton D. R. , Fischer T. P. , Walker J. A. ( 2007 ). The He–CO2 isotope and relative abundance characteristics of geothermal fluids in El Salvador and Honduras: new constraints on volatile mass balance of the Central American Volcanic Arc . Earth and Planetary Science Letters 258 , 132 – 146 . Google Scholar CrossRef Search ADS Edmond J. M. , Huh Y. ( 2003 ). Non-steady state carbonate recycling and implications for the evolution of atmospheric PCO2 . Earth and Planetary Science Letters 216 , 125 – 139 . Google Scholar CrossRef Search ADS Evans K. A. ( 2012 ). The redox budget of subduction zones . Earth-Science Reviews 113 , 11 – 32 . Google Scholar CrossRef Search ADS Evans K. A. , Tomkins A. G. ( 2011 ). The relationship between subduction zone redox budget and arc magma fertility . Earth and Planetary Science Letters 308 , 401 – 409 . Google Scholar CrossRef Search ADS Evans K. A. , Elburg M. A. , Kamenetsky V. S. ( 2012 ). Oxidation state of subarc mantle . Geology 40 , 783 – 786 . Google Scholar CrossRef Search ADS Fisk M. R. , Giovannoni S. J. , Thorseth I. H. ( 1998 ). Alteration of oceanic volcanic glass: textural evidence of microbial activity . Science 281 , 978 – 980 . Google Scholar CrossRef Search ADS Foley S. F. ( 2011 ). A reappraisal of redox melting in the Earth’s mantle as a function of tectonic setting and time . Journal of Petrology 52 , 1363 – 1391 . Google Scholar CrossRef Search ADS Frezzotti M. L. , Selverston J. , Sharp J. , Compagnoni R. ( 2011 ). Carbonate dissolution during subduction revealed by diamond-bearing rocks from the Alps . Nature Geoscience 4 , 703 – 706 . Google Scholar CrossRef Search ADS Galvez M. E. , Beyssac O. , Benzerara K. , Menguy N. , Bernard S. , Cox S. C. ( 2012 ). Micro- and nano-textural evidence of Ti(–Ca–Fe) mobility during fluid–rock interactions in carbonaceous lawsonite-bearing rocks from New Zealand . Contributions to Mineralogy and Petrology 164 , 895 – 914 . Google Scholar CrossRef Search ADS Gonzalez C. M. , Gorczyk W. , Gerya T. V. ( 2016 ). Decarbonation of subducting slabs: Insight from petrological-thermomechanical modeling . Gondwana Research 36 , 314 – 332 . Google Scholar CrossRef Search ADS Gorman P. J. , Kerrick D. M. , Connolly J. A. D. ( 2006 ). Modeling open system metamorphic decarbonation of subducting slabs . Geochemistry, Geophysics, Geosystems 7 , Q04007 . Google Scholar CrossRef Search ADS Gillis K. M. , Robinson P. T. ( 1990 ). Patterns and processes of alteration in the lavas and dykes of the Troodos Ophiolite, Cyprus . Journal of Geophysical Research 95 , 21523 – 21548 . Google Scholar CrossRef Search ADS Groppo C. , Castelli D. ( 2010 ). Prograde P–T evolution of a lawsonite eclogite from the Monviso Meta-ophiolite (Western Alps): dehydration and redox reactions during subduction of oceanic FeTi-oxide gabbro . Journal of Petrology 51 , 2489 – 2514 . Google Scholar CrossRef Search ADS Hammouda T. ( 2003 ). High-pressure melting of carbonated eclogite and experimental constraints on carbon recycling and storage in the mantle . Earth and Planetary Science Letters 214 , 357 – 368 . Google Scholar CrossRef Search ADS Harte B. ( 2010 ). Diamond formation in the deep mantle: the record of mineral inclusions and their distribution in relation to mantle dehydration zones . Mineralogical Magazine 74 , 189 – 215 . Google Scholar CrossRef Search ADS Hayes J. M. , Waldbauer J. R. ( 2006 ). The carbon cycle and associated redox processes through time . Philosophical Transactions of the Royal Society, Series B 361 , 931 – 950 . Google Scholar CrossRef Search ADS Hermann J. , Spandler C. ( 2008 ). Sediment melts at sub-arc depths: an experimental study . Journal of Petrology 49 , 717 – 740 . Google Scholar CrossRef Search ADS Hermann J. , Spandler C. , Hack A. , Korsakov A. V. ( 2006 ). Aqueous fluids and hydrous melts in high-pressure and ultra-high pressure rocks: implications for element transfer in subduction zones . Lithos 92 , 399 – 417 . Google Scholar CrossRef Search ADS Hermann J. , Troitzsch U. , Scott D. ( 2016 ). Experimental subsolidus phase relations in the system CaCO3–CaMg(CO3)2 up to 6·5 GPa and implications for subducted marbles . Contributions to Mineralogy and Petrology 171 , 84 . Google Scholar CrossRef Search ADS Hirschmann M. , Dasgupta R. ( 2009 ). The H/C ratios of Earth’s near-surface and deep reservoirs, and consequences for deep Earth volatile cycles . Chemical Geology 262 , 4 – 16 . Google Scholar CrossRef Search ADS Javoy M. , Pineau F. , Allègre C. J. ( 1982 ). Carbon geodynamic cycle . Nature 300 , 171 – 173 . Google Scholar CrossRef Search ADS Johnston F. K. B. , Turchyn A. V. , Edmonds M. ( 2011 ). Decarbonation efficiency in subduction zones: implications for warm Cretaceous climates . Earth and Planetary Science Letters 303 , 143 – 152 . Google Scholar CrossRef Search ADS Kelemen P. B. , Manning C. E. ( 2015 ). Reevaluating carbon fluxes in subduction zones, what goes down, mostly comes up . Proceedings of the National Academy of Sciences of the USA 112 , E3997 – E4006 . Google Scholar CrossRef Search ADS Kelley K. A. , Plank T. , Ludden J. , Staudigel H. ( 2003 ). Composition of altered oceanic crust at ODP Sites 801 and 1149 . Geochemistry, Geophysics, Geosystems 4 , 8910 . Google Scholar CrossRef Search ADS Kerrick D. M. , Connolly J. A. D. ( 1998 ). Subduction of ophicarbonates and recycling of CO2 and H2O . Geology 26 , 375 – 378 . Google Scholar CrossRef Search ADS Kerrick D. M. , Connolly J. A. D. ( 2001 ). Metamorphic devolatilization of subducted oceanic metabasalts: implications for seismicity, arc magmatism and volatile recycling . Earth and Planetary Science Letters 189 , 19 – 29 . Google Scholar CrossRef Search ADS Kessel R. , Ulmer P. , Pettke T. , Schmidt M. W. , Thompson A. B. ( 2005 ). The water–basalt system at 4 to 6 GPa: Phase relations and second critical endpoint in a K-free eclogite at 700 to 1400°C . Earth and Planetary Science Letters 237 , 873 – 892 . Google Scholar CrossRef Search ADS Kiseeva E. S. , Yaxley G. M. , Hermann J. , Litasov K. D. , Rosenthal A. , Kamenetsky V. S. ( 2012 ). An experimental study of carbonated eclogite at 3.5–5.5 GPa—implications for silicate and carbonate metasomatism in the cratonic mantle . Journal of Petrology 53 , 727 – 759 . Google Scholar CrossRef Search ADS Kono Y. , Kenney-Benson C. , Hummer D. , Ohfuji H. , Park C. , Shen G. , Wang Y. , Kavner A. , Manning C. E. ( 2014 ). Ultralow viscosity of carbonate melts at high pressures . Nature Communications 5 , 5091 . Google Scholar CrossRef Search ADS Lassiter J. C. ( 2003 ). Rhenium volatility in subaerial lavas: constraints from subaerial and submarine portions of the HSDP-2 Mauna Kea drillcore . Earth and Planetary Science Letters 214 , 311 . Google Scholar CrossRef Search ADS Martin L. H. J. , Schmidt M. W. , Mattsson H. B. , Guenther D. ( 2013 ). Element partitioning between immiscible carbonatite and silicate melts for dry and H2O-bearing systems at 1–3 GPa . Journal of Petrology 54 , 2301 – 2338 . Google Scholar CrossRef Search ADS Marty B. , Tolstikhin I. N. ( 1998 ). CO2 fluxes from mid-ocean ridges, arcs and plumes . Chemical Geology 145 , 233 – 248 . Google Scholar CrossRef Search ADS Mattison C. G. , Tsujimori T. , Zhang R. Y. , Liou L. G. ( 2004 ). Epidote-rich talc-kyanite-phengite eclogites, Sulu terranes, eastern China: P-T-fO2 estimates and the significance of the epidote-talc assemblage in eclogite . American Mineralogist 89 , 1772 – 1783 . Google Scholar CrossRef Search ADS McCubbin F. M. , Sverjensky D. A. , Steele A. , Mysen B. O. ( 2014 ). In-situ characterization of oxalic acid breakdown at elevated P and T: implications for organic C–O–H fluid sources in petrologic experiments . American Mineralogist 99 , 2258 – 2271 . Google Scholar CrossRef Search ADS Molina J. F. , Poli S. ( 2000 ). Carbonate stability and fluid composition in subducted oceanic crust: an experimental study on H2O–CO2-bearing basalts . Earth and Planetary Science Letters 176 , 295 – 310 . Google Scholar CrossRef Search ADS Morgan G. B. , Chou I.-M. , Pasteris J. D. ( 1992 ). Speciation in experimental C–O–H fluids produced by the thermal dissociation of oxalic acid dihydrate . Geochimica et Cosmochimica Acta 56 , 281 – 294 . Google Scholar CrossRef Search ADS Morimoto N. ( 1988 ). Nomenclature of pyroxenes . Mineralogy and Petrology 39 , 55 – 76 . Google Scholar CrossRef Search ADS Morris J. D. , Leeman W. P. , Tera F. ( 1990 ). The subducted component in island arc lavas: constraints from Be isotopes and B–Be systematics . Nature 344 , 31 – 36 . Google Scholar CrossRef Search ADS Parkinson I. J. , Arculus R. J. ( 1999 ). The redox state of subduction zones: insights from arc-peridotites . Chemical Geology 160 , 409 – 423 . Google Scholar CrossRef Search ADS Pawley A. R. , Holloway J. R. ( 1993 ). Water sources for subduction zone volcanism: new experimental constraints . Science 260 , 664 – 667 . Google Scholar CrossRef Search ADS Peacock S. M. ( 1996 ). Thermal and petrologic structure of subductions zones. In: Bebout G. E. , Scholl D. W. , Kirby S. H. , Platt J. P. (eds) Subduction: Top to Bottom. American Geophysical Union, Geophysical Monograph 96 , 119 – 133 . Plank T. , Cooper L. B. , Manning C. E. ( 2009 ). Emerging geothermometers for estimating slab surface temperatures . Nature Geosciences 2 , 611 – 615 . Google Scholar CrossRef Search ADS Poli S. ( 2015 ). Carbon mobilized at shallow depths in subduction zones by carbonatitic liquids . Nature Geoscience 8 , 633 – 636 . Google Scholar CrossRef Search ADS Poli S. , Schmidt M. W. ( 1995 ). H2O transport and release in subduction zones: experimental constraints on basaltic and andesitic system . Journal of Geophysical Research 100 , 22299 – 222314 . Google Scholar CrossRef Search ADS Poli S. , Franzolin E. , Fumagalli P. , Crottini A. ( 2009 ). The transport of carbon and hydrogen in subducted oceanic crust: an experimental study to 5 GPa . Earth and Planetary Science Letters 278 , 350 – 360 . Google Scholar CrossRef Search ADS Rohrbach A. , Schmidt M. W. ( 2011 ). Redox freezing and melting in the Earth’s deep mantle resulting from carbon–iron redox coupling . Nature 472 , 209 – 212 . Google Scholar CrossRef Search ADS Rowe M. C. , Kent A. J. R. , Nielsen R. L. ( 2009 ). Subduction influence on oxygen fugacity and trace and volatile elements in basalts across the Cascade volcanic arc . Journal of Petrology 50 , 61 – 91 . Google Scholar CrossRef Search ADS Sano Y. , Marty B. ( 1995 ). Origin of carbon in fumarolic gas from island arcs . Chemical Geology 119 , 265 – 274 . Google Scholar CrossRef Search ADS Schmidt M. W. , Poli S. ( 1998 ). Experimentally based water budgets for dehydrating slabs and consequences for arc magma generation . Earth and Planetary Science Letters 163 , 361 – 379 . Google Scholar CrossRef Search ADS Schmidt M. W. , Vielzeuf D. , Auzanneau E. ( 2004 ). Melting and dissolution of subducting crust: the key role of white micas . Earth and Planetary Science Letters 228 , 65 – 84 . Google Scholar CrossRef Search ADS Sciurto P. F. , Ottonello G. ( 1995 ). Water–rock interaction on Zabargad Island, Red Sea—a case study: I. Application of the concept of local equilibrium . Geochimica et Cosmochimica Acta 59 , 2187 – 2206 . Google Scholar CrossRef Search ADS Shaw A. M. , Hilton D. R. , Fischer T. P. , Walker J. A. , Alvarado G. E. ( 2003 ). Contrasting He–C relationships in Nicaragua and Costa Rica: insights into C cycling through subduction zones . Earth and Planetary Science Letters 214 , 499 – 513 . Google Scholar CrossRef Search ADS Simakin A. G. , Salova T. P. , Bondarenko G. V. ( 2012 ). Experimental study of magmatic melt oxidation by CO2 . Petrology 20 , 593 – 606 . Google Scholar CrossRef Search ADS Skora S. , Blundy J. D. , Brooker R. A. , Green E. C. R. , de Hoog J. C. M. , Connolly J. A. D. ( 2015 ). Hydrous phase relations and trace element partitioning behaviour in calcareous sediments at subduction-zone conditions . Journal of Petrology 56 , 953 – 980 . Google Scholar CrossRef Search ADS Smit M. A. , Bröcker M. , Scherer E. E. ( 2008 ). Aragonite and magnesite in eclogites from the Jæren nappe, SW Norway: disequilibrium in the system CaCO3–MgCO3 and petrological implications . Journal of Metamorphic Geology 26 , 959 – 979 . Google Scholar CrossRef Search ADS Spandler C. , Pettke T. , Rubatto D. ( 2011 ). Internal and external fluid sources for eclogite-facies veins in the Monviso Meta-ophiolite, Western Alps: implications for fluid flow in subduction zone . Journal of Petrology 52 , 1207 – 1236 . Google Scholar CrossRef Search ADS Staudigel H. , Hart S. R. ( 1983 ). Alteration of basaltic glass: Mechanisms and significance for the oceanic crust-seawater budget . Geochimica et Cosmochimica Acta 47 , 337 – 350 . Google Scholar CrossRef Search ADS Stakes D. S. , O’Neil J. R. ( 1982 ). Mineralogy and stable isotope geochemistry of hydrothermally altered oceanic rocks . Earth and Planetary Science Letters 57 , 285 – 304 . Google Scholar CrossRef Search ADS Staudigel H. ( 2003 ). Hydrothermal alteration processes in the oceanic crust. In: Rudnick R. L. (ed.) The Crust . Elsevier , pp. 511 – 535 . Staudigel H. , Hart S. R. , Richardson S. H. ( 1981 ). Alteration of the oceanic crust: processes and timing . Earth and Planetary Science Letters 52 , 311 – 327 . Google Scholar CrossRef Search ADS Staudigel H. , Hart S. R. , Schmincke H.-U. , Smith B. M. ( 1989 ). Cretaceous ocean crust at DSDP Sites 417 and 418: Carbon uptake from weathering versus loss by magmatic outgassing . Geochimica et Cosmochimica Acta 53 , 3091 – 3094 . Google Scholar CrossRef Search ADS Staudigel H. , Plank T. , White B. , Schmincke H.-U. ( 1996 ). Geochemical fluxes during seafloor alteration of the basaltic upper oceanic crust: DSDP Sites 417 and 418. In: Bebout E. , Scholl W. , Kirby H. , Platt P. (eds) Subduction Top to Bottom. American Geophysical Union, Geophysical Monograph 96 , 19 – 38 . Sverjensky D. A. , Stagno V. , Huang F. ( 2014 ). Important role for organic carbon in subduction-zone fluids in the deep carbon cycle . Nature Geoscience 7 , 909 – 913 . Google Scholar CrossRef Search ADS Syracuse E. M. , van Keken P. E. , Abers G. A. ( 2010 ). The global range of subduction zone thermal models . Physics of the Earth and Planetary Interiors 183 , 73 – 90 . Google Scholar CrossRef Search ADS Taylor W. R. , Foley S. F. ( 1989 ). Improved oxygen-buffering techniques for C–O–H fluid-saturated experiments at high pressure . Journal of Geophysical Research 94 , 4146 – 4158 . Google Scholar CrossRef Search ADS Thomsen T. B. , Schmidt M. W. ( 2008 ). Melting of carbonated pelites at 2.5-5.0 GPa, silicate-carbonatite liquid immiscibility, and potassium-carbon metasomatism of the mantle . Earth and Planetary Science Letters 267 , 17 – 31 . Google Scholar CrossRef Search ADS Tiraboschi C. , Tumiati S. , Recchia S. , Miozzi F. , Poli S. ( 2016 ). Quantitative analysis of COH fluids synthesized at HP–HT conditions: an optimized methodology to measure volatiles in experimental capsules . Geofluids 16 , 841 – 855 . Google Scholar CrossRef Search ADS Tsuno K. , Dasgupta R. , Danielson L. , Righter K. ( 2012 ). Flux of carbonate melt from deeply subducted pelitic sediments: Geophysical and geochemical implications for the source of Central American volcanic arc . Geophysical Research Letters 39 , L16307 . Google Scholar CrossRef Search ADS Tumiati S. , Godard G. , Martin S. , Malaspina N. , Poli S. ( 2015 ). Ultra-oxidized rocks in subduction melanges? Decoupling between oxygen fugacity and oxygen availability in a Mn-rich metasomatic environment . Lithos 226 , 116 – 130 . Google Scholar CrossRef Search ADS van Keken P. E. , Kiefer B. , Peacock S. M. ( 2002 ). High-resolution models of subduction zones: Implications for mineral dehydration reactions and the transport of water into the deep mantle . Geochemistry, Geophysics, Geosystems 3 , 1–20 . Google Scholar CrossRef Search ADS Varekamp J. C. , Kreulen R. , Poorter R. P. E. , Van Bergen M. J. ( 1992 ). Carbon sources in arc volcanism with implications for the carbon cycle . Terra Nova 140 , 217 – 240 . Wallace P. J. ( 2005 ). Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data . Journal of Volcanology and Geothermal Research 140 , 217 – 240 . Google Scholar CrossRef Search ADS Wenzel T. , Baumgartner L. P. , Brogmann G. E. , Konnikov E. G. , Kislov E. V. ( 2002 ). Partial melting and assimilation of dolomitic xenoliths by mafic magma: the Ioko-Dovyren intrusion (North Baikal region, Russia) . Journal of Petrology 43 , 2049 – 2074 . Google Scholar CrossRef Search ADS Wyllie P. J. , Tuttle O. F. ( 1959 ). Melting of calcite in the presence of water . American Mineralogist 44 , 453 – 461 . Yaxley G. M. , Green D. H. ( 1994 ). Experimental demonstration of refractory carbonate-bearing eclogite and siliceous melt in the subduction regime . Earth and Planetary Science Letters 128 , 313 – 325 . Google Scholar CrossRef Search ADS Zhang R. Y. , Liou J. G. , Ernst W. G. ( 2009 ). The Dabie–Sulu continental collision zone: a comprehensive review . Gondwana Research 16 , 1 – 26 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of PetrologyOxford University Press

Published: Apr 2, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off