Experimental Constraints on Sulphur Behaviour in Subduction Zones: Implications for TTG and Adakite Production and the Global Sulphur Cycle since the Archean

JOURNAL OF PETROLOGY VOLUME 54 NUMBER 1 PAGES 183^213 2013 doi:10.1093/petrology/egs067 Experimental Constraints on Sulphur Behaviour in Subduction Zones: Implications for TTG and Adakite Production and the Global Sulphur Cycle since the Archean UNIVERSITE D’ORLE¤ANS, ISTO, UMR 7327, 45071, ORLE¤ANS, FRANCE CNRS/INSU, ISTO, UMR 7327, 45071 ORLE¤ANS, FRANCE 1 2 3 BRGM, ISTO, UMR 7327, 45071 ORLE¤ANS, FRANCE RECEIVED DECEMBER 10, 2010; ACCEPTED AUGUST 31, 2012 ADVANCE ACCESS PUBLICATION OCTOBER 25, 2012 We performed crystallization experiments at 2^3 GPa at 700^ 9508C on basaltic and pelitic lithologies with added water and sulphur to constrain the factors controlling sulphur behaviour in subduction zones and how it may have varied through geological time. The resulting hydrous silicic melts have up to 20 times more dissolved sulphur (up to 1wt %) than at 0·2^0·4 GPa, when moderately oxidized conditions prevail. Such high solubilities appear to result from the combined effects of enhanced solubility of water in high-pressure silicate melts (10^20 wt % H2O), which acts to decrease silica activity, and oxidizing conditions. The results confirm previous findings that high sulphur contents in silicate melts do not necessarily require iron-rich compositions, suggesting instead that sulphur^ water complexes play a fundamental role in sulphur dissolution mechanisms in iron-poor silicic melts, in agreement with recent spectroscopic data. The experimental melts reproduce Phanerozoic slab-derived magmas, in particular their distinct Ca- and Mg-rich composition. The results also show that sulphur increases the degree of melting of basalt lithologies. Hence, we suggest that subducted slabs will preferentially melt where sulphur is present in abundance and that the variability in arc magma sulphur output reflects, in part, the vagaries of sulphur distribution in the slab source. In contrast, comparison with the composition of Archean felsic rocks suggests that, in the early Earth, much less sulphur was present in subducted slabs, in agreement with a number of independent lines of evidence showing that the Archean ocean, hence the hydrothermally altered subducted Archean oceanic crust, was considerably poorer in The behaviour of sulphur in subduction zones remains largely unknown. A number of recent eruptions have illustrated that arc magmas are important contributors of sulphur to the atmosphere (Robock, 2000) with restored bulk sulphur contents prior to eruption (including those with a strong slab melt imprint) of c. 0·5 wt %, reaching several wt % in some instances (Scaillet et al., 2003). Experiments and calculations have shown that most of the S is in the fluid phase (Scaillet et al., 1998; Keppler, 1999; Burgisser & Scaillet, 2007; Burgisser et al., 2008), which exsolves prior to eruption. High sulphur concentrations (up to 6000 ppm) in silicate-melt inclusions hosted in olivine from arc harzburgitic xenoliths and primitive basaltic lava samples (Me¤trich et al., 1999; De Hoog et al., 2001; Cervantes & Wallace, 2003) indicate that arc mantle may be significantly enriched in sulphur (up to 500 ppm) *Corresponding author. Present address: CNRS/INSU, ISTO, UMR 7327, 45071, Orle¤ans, France. E-mail:  The Author 2012. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@ oup.com sulphur than at present. Volcanic degassing of sulphur was thus probably much weaker during the Archean than in Proterozoic^ Phanerozoic times. KEY WORDS: adakite; magma; partial melting; sulphide; subduction; thermodynamics; igneous petrology; experimental petrology I N T RO D U C T I O N GAE«LLE PROUTEAU1,2,3 AND BRUNO SCAILLET1,2,3* JOURNAL OF PETROLOGY VOLUME 54 JANUARY 2013 and/or stress-dependent rheology (e.g. van Keken et al., 2002; Kelemen et al., 2003), point to top slab temperatures high enough to allow fluid-saturated melting of the subducted slab. In support of this are recent findings based on trace element solubilities in high-pressure fluids and melts that point to the temperatures at the top of the slab high enough to allow its melting (i.e. 47008C; Plank et al., 2009). Additionally, sediment partial melting is required for critical elements such as Th and Be to be efficiently transferred into the mantle wedge (Johnson & Plank, 1999; Hermann & Spandler, 2008), which calls for partial melting of subducted slabs to be more common than generally thought. With respect to the first limitation, no experimental study has yet explored the solubility of sulphur in hydrous silicic melt at high pressures. Available experimental constraints on sulphur solubility in silicate melts at high pressures concern mostly dry basaltic compositions (e.g. Wendlant, 1982; Mavrogenes & O’Neill, 1999; Holzeid & Grove, 2002; Jugo et al., 2005), which are not appropriate to most subduction zone environments. Within this context we have investigated the sulphur solubility in melts produced from partial melting of basalt and sediment, under P^T^fO2 conditions relevant to subduction zones. Our approach consisted of adding sulphur to basalt^H2O or sediment^H2O charges, whose melting behaviour without sulphur had been previously determined (Prouteau et al., 2001; Hermann & Spandler, 2008). We explored the effect of oxygen fugacity (fO2), owing to its influence on sulphur behaviour in silicate melts (Carroll & Rutherford, 1987; Luhr, 1990; Carroll & Webster, 1994; Scaillet & Evans, 1999; O’Neill & Mavrogenes, 2002; Jugo et al., 2005), from around NNO þ1 (one log unit above the solid buffer nickel^nickel oxide) to below NNO ^ 2. We explored a range of 1^4 wt % bulk sulphur content to determine the maximum sulphur content in silicate melts under these conditions. It should be noted that we do not argue that the oceanic crust is everywhere enriched at the per cent level in sulphur, but rather that such concentrations may locally be reached and exert a first-order control on the observed variability in the sulphur output of arc magmas (e.g. Scaillet et al., 2003). We investigated a pressure and temperature range of 2^3 GPa and 700^9508C, respectively. Such conditions include those of the top of the subducted slab beneath the majority of active volcanic arcs and are thus supposed to mimic slab melting (Defant & Drummond, 1990; Schmidt & Poli, 1998; Kelemen et al., 2003; Hermann & Spandler, 2008). We have not explored higher pressures primarily to avoid approaching too closely the second critical end point (Bureau & Keppler, 1999; Kessel et al., 2005a, 2005b; Klimm et al., 2008), which for the basalt^H2O system is considered to lie between 5 and 6 GPa (Kessel et al. 2005b). The experimental pressures are thus sufficiently high to approach slab melting conditions, but also low enough to avoid H2O contents of 184 relative to mid-ocean ridge basalt (MORB) mantle (c. 200 ppm), following metasomatism by an H2O-rich agent with up to several wt % su (...truncated)