Here we report new data on the sulfur isotopic compositions (δ34S) of fumarolic and plume gases collected at Mount Etna volcano during 2008–2009. While low-temperature fumaroles are affected by postmagmatic processes that modify the pristine isotopic signature, high-temperature and plume gases allow establishment of aδ34S range of ∼0 ± 1‰ for magmatic SO2. We compared our data with those from S dissolved in primitive melt inclusions from 2002 lava and in whole rocks that erupted during the past two thousand years. Such a comparison revealed that δ34S is systematically lower for magmatic gases than for sulfur dissolved in the melt. We modeled how isotopic fractionation due to magma degassing process may vary δ34S value in both the melt and gaseous phases. This modeling required assessment of the fractionation factor (αgas-melt). The most recent measurements on the oxidation state of sulfur in basaltic melt inclusions indicate that nearly all S is dissolved as sulfate (S6+), which would be possible in oxidized magmatic systems (ΔNNO ≥ 1). Under these conditions the exsolved gaseous phase is depleted with respect to the melt and the proposed model fits both gas and melt data, and constrains the Etnean magmatic δ34S to 1.0 ± 1.5‰. It is remarkable that the assessed redox conditions—which are significantly more oxidizing than previously thought—are able to explain why the dominant sulfur species measured in the Etnean plume is SO2.
 Mount Etna appeared about 500,000 years ago near to the collision boundary between African and European plates, and is one of the most active volcanoes worldwide. It is frequently accompanied by explosive (mainly ash emissions, Strombolian activity, and fire fountains) and/or effusive eruptions. One of its prominent features is the huge amounts of H2O, CO2, and SO2 from the summit craters degassed in the atmosphere, with an estimated average total flux of ∼21,000 t·day−1 during 2005–2006 [Aiuppa et al., 2008]. Despite the SO2 flux accounting for only a small fraction (<17%) of the total outgassing budget [Aiuppa et al., 2008], sulfur degassing has great implications in the evolution of magma, as a potential trigger of eruptions, and in estimating the magma supply at depth [Oppenheimer et al., 2011; Steffke et al., 2011, and references therein]. An investigation of the sulfur content in glass inclusions hosted in olivines from lava that erupted at Mount Etna during 2002 demonstrated that it is fully dissolved in the melt (0.28–0.34 wt%) at up to ∼1500 bar, and that it is lost from magma during its ascent at lower pressures and shallower depths [Spilliaert et al., 2006]. This means that sulfur degassing beneath Mount Etna becomes appreciable only at shallow portions of the plumbing system (less than 5 km below the volcano summit), and thus variations of sulfur content in crater gases may represent a useful indicator of magma approaching the surface.
 Mount Etna crater gases were investigated by Martelli et al. , Liotta et al.  and Paonita et al.  for their chemical composition, isotopic ratios of C(CO2), and noble gases ratios. Gas emissions contain two sulfur species (mainly SO2 and some H2S) depending on the pressure, temperature, and redox conditions of the degassing melt [Aiuppa et al., 2005] and on the occurrence of secondary processes affecting the composition of magmatic gases [Liotta et al., 2010]. Particularly, CO2-dominated low-temperature (about 90°C) fumaroles provide SO2 and H2S in comparable proportions, with SO2/H2S ratios ranging between 1.5 and 4. On the other hand, high-temperature (about 400°C) fumaroles and plume gases are poorly or not affected by hydrothermal fractionation processes [Liotta et al., 2010]. In addition, several investigations of Etnean plume gases have demonstrated that the SO2/H2S ratio varies from 20 to almost 215 [Aiuppa et al., 2005, 2011; Oppenheimer et al., 2011, and references therein].
 There are some data of sulfur isotopes at Mount Etna volcano. Most of them relate to S dissolved in whole rocks that erupted during the past two thousand years [Gambardella, 2000] and in primitive melt inclusions (MIs) hosted in olivines from lava that erupted in 2002 [Allard et al., 2006; Spilliaert, 2006]. Sulfur isotope were analyzed in SO2discharged from high-temperature (780–1000°C) fumaroles collected between 1975 and 1979 [Allard, 1978, 1983, 1986], and as dissolved sulfate in Etnean groundwater sampled in 2002 [D'Alessandro et al., 2003]. Also, volcanic SO2 emitted from the Voragine crater during 2004 was analyzed for the determination of the 34S/32S ratio by using filter packs [Mather et al., 2008].
 Herein we describe new measurements of the δ34S ratio at Mount Etna volcano carried out on gas samples discharged from both low- and high-temperature fumaroles and on plume gas from the central craters that were collected during the 2008–2009 eruptive period. Our data are then compared with all of the data available in the literature from the same area in order to assess the sulfur isotopic signature of Etnean melts and to investigate secondary processes able to modify the pristine magmaticδ34S ratio.
2. Sampling and Analyses
 A new subterminal eruption occurred at Mount Etna from May 13, 2008 to July 6, 2009. Its onset was characterized by fire fountains, falling tephra, and lava effusion toward Valle del Bove that continued vigorously through late-July 2008, after which all explosive (Strombolian) activity ceased and lava emissions decreased gradually [Bonaccorso et al., 2011]. Two sampling campaigns (in October 2008 and July 2009) were performed in the crater area during this eruption. Gases were collected from two fumaroles [F8-F9 (temperature ∼90°C) and F4 (temperature ∼400°C); hereafter referred to as LT and HT, respectively] and from the plume of the central craters (Figure 1). LT fumaroles were sampled by using Giggenbach bottles filled with a 4 M NaOH solution in order to dissolve the acidic species [Giggenbach, 1975]. Due to the high air contamination (N2 > 60%), HT fumarole was sampled in a pre-evacuated glass bottle having a volume of 4 l, into which 100 ml of 1 M NaOH solution was injected after collection, thus allowing total dissolution of the acidic gases and their detection by ion chromatography. Plume gases were collected through a bubbler (Figure 1) powered by a battery and regulated through a flowmeter (set to 1 l·min−1) that was connected to a glass bottle constructed to allow the gas to dissolve in an alkaline solution (1 M NaOH). This homemade system was placed for about 1 h at the eastern edge of the central craters, because of the wind conditions of the sampling day, in an area free of soil gas emissions. All samples were processed (oxidized and acidified) and analyzed for sulfur isotopic composition at the NERC Isotope Geosciences Laboratory, British Geological Survey according to Carmody et al. . Results are given as δ34S values on the V-CDT (Vienna Canyon Diablo Troilite) scale inTable 1, together with the analytical uncertainties. It should be remembered that the isotopic data reported here refer to the composition of total sulfur (δ34SΣS), even in the case of LT fumaroles that discharge slightly higher amount of SO2 with respect to H2S [Liotta et al., 2010].
Table 1. Data Set of Available Sulfur Isotopic Composition at Mount Etna Volcanoa
[S]T indicates sulfur concentration in the samples, expressed as ppm for whole rock and melt inclusions, while as %vol for dry gases and as μmoles for filter packs; n.a. = not available.
 Data of the sulfur isotopic composition are presented here relative to the type of gas emission collected during the two field campaigns (see Table 1). The δ34S values ranged between +3.1‰ and +3.8‰ for the LT fumaroles and between −0.9‰ and −0.2‰ for the HT fumarole (Figure 2). The ratios for the samples collected during 2009 (both LT and HT gases) were comparable or slightly lower than those collected during 2008. Plume gases, which were investigated for sulfur isotopes only in 2009, show a δ34S value of +0.5‰ (Figure 2). According to Liotta et al. , the HT fumarole shows minor or negligible modifications due to secondary processes (e.g., dissolution of SO2 and H2S, and deposition of elemental sulfur), and therefore its data accurately reflects the isotopic composition of SO2 outgassed from magma at Mount Etna volcano. This inference is substantiated by the similarity of the δ34S values for HT and plume gases, based on the reasonable assumption that the latter provide the isotopic composition of magmatic sulfur. Therefore, we infer that the mean δ34S value of −0.2‰ (SD = 0.7‰) measured at the HT fumarole and in plume gases is representative of the pristine SO2 degassed from the magma during 2008–2009.
Liotta et al.  suggested that LT gases are affected by postmagmatic processes during their ascent to the surface, which would have also modified the δ34S ratios of pristine magmatic fluids, resulting in increased δ34S ratios for the LT fumaroles; this is discussed in section 4.1.
3.1. Previous Data of δ34S at Etna
 High-temperature volcanic gases (950–1000°C) were collected in 1975 and 1976 from two eruptive fractures on the northeast flank of the volcano, and fumaroles (780–800°C) sampled in 1979 from the northeast and southeast craters [Allard, 1978, 1983, 1986]. The 1975 volcanic gases showed δ34S values between +2.1‰ and +3.1‰, while those of 1976 ranged between +0.6‰ and +1.4‰. The fumaroles sampled in 1979 varied over a narrower range, from +0.7‰ to +1‰ (Figure 2). Allard [1978, 1983] stated that the 1975 data were affected by secondary processes and/or sampling problems that would have increased the amount of heavy sulfur. Therefore, in order to constrain the magmatic δ34S, we only take into account the fumarolic values of 1979 (note that the data from 1976—with the exception of a single value of δ34S = +1.4‰—overlap those of 1979).
 Sulfur isotopes investigated in 2002 in dissolved sulfates of Etnean groundwater displayed a wide isotopic range from −0.4‰ to +28.5‰ [D'Alessandro et al., 2003]. The authors interpreted these data as the result of a mixing between a marine (+21‰) and a magmatic source, with the δ34S values >21‰ probably affected by fractionation processes in groundwater. The lowest δ34S value (i.e., −0.4‰) was suggested to reflect the typical magmatic SO2 [D'Alessandro et al., 2003]. This inference is also supported by the S/Cl ratio dissolved in water being negatively correlated with the δ34S value of sulfate, and matching the highest value of S/Cl ∼3 [D'Alessandro et al., 2003]. Such an S/Cl ratio mirrors that of the plume emitted from the craters [Polacci et al., 2009]. During 2004 the isotopic composition of volcanic sulfur outgassed from Voragine crater was investigated by using filter packs, which revealed unexpectedly high δ34S ratios between +5.5‰ and +6.6‰ (Figure 2) [Mather et al., 2008]. Those authors found that fractionation of sulfur isotopes occurs when SO2from concentrated volcanic gas plumes is collected on alkali-impregnated filter papers. In view of what above stated, data ofδ34S from dissolved sulfate and filter packs cannot be considered representative of magmatic sulfur and are not discussed further.
 Additional studies of δ34S at Mount Etna have focused on the isotopic compositions of sulfur dissolved in MIs and/or lavas. Gambardella  investigated S retained in whole rocks of some historical and modern lavas erupted during the past two thousand years; they showed δ34S values between +2‰ and +4.1‰ (Figure 2). More recent preliminary analyses of primitive MIs hosted in olivines from 2002 lava were carried out using SIMS techniques [Allard et al., 2006; Spilliaert, 2006], which revealed δ34S ratios that were even more variable, ranging from +0.5‰ to +4.7‰ (Figure 2).
 The available data of the S isotopic composition previously measured at Mount Etna can be compared to our data. We recall that δ34S from volcanic gases collected in 1979 showed values generally comprised between +0.6‰ and +1‰ [Allard, 1978, 1983, 1986]. When including these values in the magmatic range given by this study (δ34S = −0.2‰ ± 0.7‰), we can assume that a δ34S signature of ∼0 ± 1‰ reflects gaseous SO2 discharged from Etnean melts.
 Differently, data from sulfur dissolved in lavas displayed values of δ34S systematically higher than the above-estimated range of magmatic gases (HT fumarole and plume), while those from MIs cover both gas and whole-rock data ranges. It should be noted thatδ34S values measured in primitive MIs should mirror the isotopic signature of dissolved S. Despite the total S content measured by Spilliaert  falling within a very narrow range (0.27–0.34 wt%; see Table 1), which is typical of Etnean melts that are not yet S-degassed [Spilliaert et al., 2006, and references therein], the variability of δ34S ratios is doubtfully large and is not related to S contents. If we consider that the δ34S values of fumaroles and plume gases spanning a range of 2‰ represent different periods of degassing, and that a similar span also involves whole rocks from eruptions on a longer time period [Gambardella, 2000], the MIs data appear too variable to be representative of the magma source, especially considering that MIs belong to a single eruption. Therefore, the large isotope range covered by these data means that the S isotopic signature of the source of Etnean magmas remains unknown.
4.1. Secondary Processes Affecting δ34S in Fumarolic Gases
 LT fumaroles show δ34S values that are systematically higher than those of HT gases, and also slightly exceed those of volcanic gases collected during 1975–1979 by Allard . We deduce that such values do not reflect the pristine isotopic composition of S released by magma (δ34S ∼0 ± 1‰), but rather are moderately affected by postmagmatic processes. In fact, after S has been degassed from magma it participates in several chemical reactions also involving aqueous phases when the pressure and temperature conditions allow liquid water to form. The sulfur isotopic ratio can then be modified by dissolution of SO2 and H2S, by isotopic exchange with aqueous and solid phases, and by deposition and/or remobilization of sulfur from sublimates.
 At temperatures below 200°C, rising volcanic gases can reach saturation with respect to the deposition of elemental sulfur [Giggenbach, 1987]. Sulfur deposition can be observed in many volcanic systems containing fumarole fields down to boiling temperature [Giggenbach et al., 1990, and references therein]. Since elemental sulfur is enriched in the light isotope 32S, this process induces a preferential removal of 32S from the gas phase, with the consequence that the residual gas (mainly SO2) will be enriched in 34S [Sakai, 1957; Grinenko and Thode, 1970; Giggenbach et al., 1990]. Even though such a process is commonly believed to be the main one leading to 34S enrichment in the gaseous phase, at Etnean fumaroles its solid deposition probably occurs to a low extent [Liotta et al., 2010], so we cannot exclude that additional processes may lead to sulfur fractionation (both chemical and isotope). Among these, sulfur dissolution in high-temperature aqueous solution and hydrothermal contributions from external sources of heavy sulfur (i.e., leaching of lavas) could play an important role. These processes would be in agreement with the hydrothermal model ofLiotta et al.  that predicts advanced interaction between volcanic gases and hydrothermal solutions. The authors found comparable contents of SO2 and H2S in LT fumaroles, by showing that cooling of magmatic gases induces the rapid conversion of SO2 into H2S. Because SO2/H2S ratio measured in plume gases varies from 20 to almost 215 [Aiuppa et al., 2005, 2011; Oppenheimer et al., 2011, and references therein], they considered H2S measured in LT fumaroles as due to the above described process. Nevertheless, it must be noted that this process would not have any effect on δ34S of total sulfur, that we measured in our samples, and cannot explain the observed 34S enrichment. Thus, further investigations are necessary to better constrain the secondary processes that may affect sulfur isotopes in LT gases.
where X and Y are the molar fractions of sulfur dissolved in the melt as sulfate and sulfide, respectively, and A and B are the molar fractions of H2S and SO2 in the gas, respectively. Equation (1) assumes thus that all sulfur occurs in silicate melts as both reduced (S2−) and oxidized (S6+) species, while SO2 and H2S are the sole sulfur-bearing molecules in the gas phase. The H2S-SO42− and S2−-SO42− fractionation factors in the melt are from Miyoshi et al. , while the H2S-SO2 values are from Ohmoto and Rye .
 The application of equation (1) requires knowledge of S speciation in both the gas and melt. For the melt, we generally use the relative abundances of S2− and S6+ measured in MIs that, if available, can be considered to be representative of the magmatic conditions. In contrast, constraining the SO2/H2S ratio of the gaseous phase coexisting with magma at depth requires a direct measurement, which cannot be performed; we therefore need to know both the fO2 and fH2O in order to recalculate the ratio. fO2 can be obtained from S6+/Stot measurements by applying available speciation models. The most applied models have been developed by Wallace and Carmichael  and Jugo et al. [2005, 2010]: the former was based on the work of Carroll and Rutherford  but investigated a wider range of compositions and temperatures based on submarine basaltic glasses, while Jugo et al. [2005, 2010]updated the available data set with experimental data obtained by both electron microprobe (EMPA) and X-ray absorption (XANES) methods especially at high oxygen fugacities, and they provided new empirical calibrations. Here we neglect thefO2-S6+/Stot relationship given by Marini et al. —based on the models by Moretti et al.  and Moretti and Papale —as being entirely theoretical and not based on the available experimental measurements, which produces results that differ markedly from those of the semi-empirical models. The consequences of using it are described insection 4.3.
Figure 3a shows how the S6+/Stot ratio in the basaltic melt varies as a function of fO2 (expressed as ΔNNO) according to the models developed by Wallace and Carmichael  and Jugo et al. [2005, 2010]. Figure 3b indicates how the molar fraction of SO2 in the gas changes as a function of ΔNNO at different fH2O values. Once the sulfur speciation ratio in the melt is defined and realistic fH2O values are evaluated, the SO2 fraction in the gas phase can be computed within the oxygen fugacity range found by the empirical models of S speciation in the melt.
 There are few direct measurements of S6+/Stot in Etnean melts. Métrich and Clocchiatti  investigated the S content (0.27–0.34 wt%) and the relative proportions of S2– and S6+ in primitive MIs hosted in olivines from scoria ejected during 1989–1990, and estimated that the average S6+/Stot ratio was 0.44. Following Wallace and Carmichael , they calculated a relative oxygen fugacity of ΔNNO = 0.35. Measurements of sulfur speciation in Etnean MIs by Moretti  yielded S6+/Stot ratios of up to 0.9. More recently, Métrich et al.  studied the oxidation state of sulfur in basaltic inclusions hosted in olivines from 2001 lava, and found that nearly all S is dissolved as sulfate (S6+), which is possible only in more oxidized systems. In terms of water fugacity, a model of magma degassing based on the pressure-dependent solubilities of the main degassed species (i.e., H2O, CO2, and S) showed that S is normally fully dissolved in the Etnean melts at pressures >1000 bar [Spilliaert et al., 2006]. This value can thus be assumed to be the upper limit for water fugacity during sulfur degassing.
 If the molar fraction of SO4 in the melt is 0.44 (Figure 3a), the calculated fO2 ranges in terms of ΔNNO between 0.30 and 0.62 (i.e., minimum and maximum values obtained by the speciation models). Within this range, when water fugacity is ≤1000 bar, the SO2/H2S ratio is always ≥1.1 (B ≥ 0.53; Figure 3b). As a consequence, the application of equation (1) yields ln(αgas-melt) values ranging between −0.36‰ and +0.58‰.
 As an alternative we consider an S6+/Stot value of ≥0.95, based on Métrich et al.  reporting S6+/Stot ∼1. In this case fO2 leads always to ΔNNO ≥ 1 (Figure 3a) and, when fH2O is ≤1000 bar, the SO2/H2S ratio is always ≥13 (B ≥ 0.9; Figure 3b). In this case the application of equation (1) yields ln(αgas-melt) values ranging from −1.55‰ to −1.24‰.
 We recall that the estimated range of δ34S values for HT crater gases (0 ± 1‰) was about 2‰ lower than that measured for most of the dissolved S. These data would imply an exsolved gas phase depleted in 34S with respect to the melt, which contrasts with the αgas-melt value calculated by using S6+/Stot = 0.44. Even considering the most reducing conditions (ΔNNO = 0.30) at fH2O = 1000 bar, the ln(αgas-melt) value (−0.36‰) does not explain the difference in δ34S values between gaseous and dissolved sulfur. In addition, under these conditions the expected SO2/H2S ratio in the gas phase would be 1.1, which is incompatible with the direct measurements in the plume gases of values between 20 and 215 [Aiuppa et al., 2005, 2011; Oppenheimer et al., 2011, and references therein].
 Conversely, the Etnean data set is more easily explained with ln(αgas-melt) values ranging from −1.55‰ and −1.24‰ (resulting, as explained above, by S6+/Stot ≥ 0.95). Within this range, even considering the most reducing conditions (ΔNNO = 1) at fH2O = 1000 bar, the expected SO2/H2S ratio would be 13, thus implying that most of the S is degassed from Mount Etna craters as SO2. At slightly more oxidizing conditions, the ratio essentially matches the measurements carried out at the summit craters [Aiuppa et al., 2005, 2011; Oppenheimer et al., 2011, and references therein]. An oxygen fugacity of ΔNNO ≥ 1 is also compatible with reconstructions based on the Cr-spinel composition made byKamenetsky and Clocchiatti , who extended the range of redox conditions at Etna to more oxidizing conditions (ΔNNO up to ∼1.8).
 In addition, Métrich and Clocchiatti  and Métrich et al.  measured the S6+/Stot ratio using EMPA and XANES techniques, respectively. Jugo et al.  compared these two analytical techniques and found that EMPA analysis tend to underestimate S6+/Stot ratio under oxidizing conditions in samples that are dominated by sulfate (Beermann et al.  confirmed these finding). This would further support the hypothesis that S6+/Stot ∼1 (measured by Métrich et al. ) more realistically reflects the sulfur speciation in Etnean melts, and hence below we focus on this case.
4.3. Effect of Magma Degassing on Sulfur Isotopes
 Once the redox conditions for the Etnean magmatic system at ΔNNO > 1 and consequently the estimated range of possible ln(αgas-melt) values (−1.55‰ to −1.24‰) have been defined, we can calculate the evolution of the sulfur isotopic composition of magma as degassing proceeds. Figure 4 shows the δ34S changes in the melt and gas versus the sulfur residual fraction in the melt (hereafter F) during open-system degassing, computed by using an equation for Rayleigh distillation. The sulfur content (mean = 0.33 wt%, SD = 0.03 wt%) measured in primitive MIs bySpilliaert  and Spilliaert et al. can be considered to be the initial S content in the melt, and thus MIs are plotted in the figure at F = 1. The F value for each whole-rock sample obtained byGambardella  was computed by dividing its measured S content (see Table 1) by that of the initial melt. In order to constrain the starting conditions for the S isotopic composition, we recall that our estimated δ34S range for magmatic gases was 0 ± 1‰, which reflects both the plume and HT-fumarole compositions. Nevertheless, we need to constrain the F value at which the gases were exsolved. During the same sampling period in 2008,Aiuppa et al.  measured CO2/SO2 ∼7 in gases emitted in the plume of the central craters. Following the degassing models developed to interpret such a ratio [Aiuppa et al., 2007; Shinohara et al., 2008], when calculating a residual fraction leading to CO2/SO2 ∼7 both models suggest F > 0.5. In addition, it is notable that the highest residual fractions in whole-rock samples obtained byGambardella  exhibit F ∼0.5 (Figure 4). Considering that lava samples generally reflect highly degassed conditions (as indicated by very low H2O contents), we can reasonably assume that gases collected at the craters reflect weaker degassing. Thus, we assume 0.5 to be the lowest F value at which our sampled gases are exsolved. This means that the sampled gases could be exsolved at F values ranging from 0.5 to 1, with the latter value indicative of pristine magmatic gas. Below we model the sulfur isotopic fractionation due to magma degassing in the following two limiting cases: δ34S = 0 ± 1‰ with F = 1 and δ34S = 0 ± 1‰ with F = 0.5 (Figures 4a and 4b, respectively).
 Within the range defined for the sulfur fractionation, open-system degassing leads to a progressive enrichment in the34S values of both melt and gaseous phases (Figure 4). At F = 1 (Figure 4a), the sulfur isotopic composition of whole-rock samples is only partially consistent with the fractionation paths. Under this condition theδ34S value of sulfur dissolved in Etnean parental magma would range from +0.3‰ to +2.5‰, and MIs exhibiting δ34S ≤ +2.5‰ would be explained by this scenario. Alternatively, in the case of F = 0.5 (Figure 4b), the parental exsolved gas would be characterized by −2.0‰ ≤ δ34S ≤ 0.1‰ and the value for sulfur dissolved in the coexisting parental magma would range from −0.5‰ to +1.5‰. These conditions would be consistent with most of the whole-rock samples but only with MI data for whichδ34S < +2‰. However, although all intermediate conditions between F = 0.5 and F = 1 are possible, research at a finer resolution would currently be too ambitious. Despite this uncertainty, we are able to constrain the magmatic δ34S value to be within the range of 1.0 ± 1.5‰. What is remarkable is that oxidizing conditions (ΔNNO ≥ 1) must dominate in the shallow plumbing system in order to explain the available data set for δ34S values, as well as sulfur speciation in both the melt and plume.
 Some of the whole-rock (i.e., 122 Before Christ (BC) samples;Table 1) and MI data have also been interpreted by Marini et al. . Those authors modeled a fractionation path during open-system degassing by assuming S6+/Stot = 0.64 in the melt, even when this ratio was measured to vary from 0.6 to 0.9 in MIs from 122 BC products [Moretti, 2002]. By using the sulfur speciation model of Moretti et al.  and Moretti and Papale , they computed an oxygen fugacity in terms of ΔNNO = −0.08 at fH2O ≈ 1000 bar. Such conditions are much more reducing than those when using the semi-empirical models inFigure 3a. Under these reducing conditions, Marini et al.  found sulfur speciation in the gaseous phase with SO2/Stot = 0.61. Our approach is comparable with theirs, but the redox conditions that they found differed markedly from ours. The calculated SO2/Stot value of 0.61 would imply the presence of ∼40% H2S in the exsolved gas phases, which seems unrealistic considering that direct measurements of S degassed from the Mount Etna craters have yielded SO2/H2S ratios between 20 and 215 [Aiuppa et al., 2005, 2011; Oppenheimer et al., 2011, and references therein]. In addition, as already noted, the redox conditions that we suggest using to interpret the entire Etnean data set (ΔNNO ≥ 1) are in better agreement with those reported by Kamenetsky and Clocchiatti  and Métrich et al. .
 The magmatic isotopic signature we defined can be discussed in terms of the source tracer, however it should be noted that sulfur isotopes are less discriminating than are other geochemical tracers (e.g., Sr, Nd, Pb, and He isotopes). In fact, δ34S values from MORB-like environments move generally within the range of ∼0 ± 2‰ [Marini et al., 2011, and references therein] and cover the δ34S values measured in basaltic glasses and volcanic gases of Kilauea volcano (from −0.8‰ to +0.9‰ [Sakai et al., 1982]), which are representative of plume source. On the other hand, arc volcanic rocks range from the previous values up to a particularly enriched value of +20‰ [Ueda and Sakai, 1984; Taylor, 1986; Marini et al., 2011, and references therein] due to recycling of marine sulfate (δ34S = +21‰) during subduction. As an example, δ34S values more positive than MORB were found in fumarolic gases collected from the subduction-related volcano of White Island (New Zealand) (δ34S ∼+3‰ [Rafter et al., 1958; Giggenbach, 1987; Marini et al., 2011]). In the Aeolian arc (Italy), 70 km north of Mount Etna, Vulcano Island displays δ34S values from −2.5‰ to +3‰ [Cortecci et al., 1996] that result from magmatic sulfur (+0.5‰ to +3‰) mixing with lighter sulfur of hydrothermal origin [Di Liberto et al., 2002]. Within the Mediterranean context, the origin of sulfur has been also investigated between 1983 and 1988 in the Solfatara fumaroles of Campi Flegrei (Southern Italy), with a mean of −0.3 ± 0.3‰ [Allard et al., 1991]. In the period 1998–2001, new measurements displayed average δ34S value of +1.7 ± 0.3‰ [Marini et al., 2011]. The magmatic systems above discussed are commonly characterized to be subduction-related.
 Comparison with the above-given ranges indicates that the sulfur isotopic ratios of parental Etnean magmas (δ34S = 1.0 ± 1.5‰) partially overlap both the MORB-likeδ34S values and those of the near subduction-related volcanisms. This would imply that primary melts could originate from a mantle source metasomatised by slab-derived fluids, as already suggested by several authors [Tonarini et al., 2001; Armienti et al., 2004, and references therein]. Nevertheless, due to the limited δ34S data still available at Etna, it is not feasible to make clear inferences about the Etnean mantle and its geodynamic features and further studies on the local magmatism are necessary.
5. Concluding Remarks
 In this work we investigated the sulfur isotopic composition measured in crater fumaroles and plume gases at Mount Etna volcano. The main results are summarized as follows:
 1. Low-temperature fumaroles are affected by postmagmatic processes that shift the S isotopic signature toward more positiveδ34S values with respect to high-temperature and plume gases, which are believed to be representative of pristine magmatic sulfur.
 2. SO2 directly exsolved from magma displayed δ34S ∼0 ± 1‰, being systematically 1–2‰ lower than that of S dissolved in Etnean melts. This difference is attributed to isotopic fractionation of sulfur between melt and gas phases during magma degassing. By assuming S6+/Stot ∼1 in the melt, as estimated from the most recent speciation measurements in Etnean basalt [Métrich et al., 2009], we calculated ln(αgas-melt) values between −1.55‰ and −1.24‰ at oxygen fugacity ΔNNO ≥ 1. Under such conditions we showed that a Rayleigh-type degassing model is able to explain almost the entire data set (MIs, rocks, and gases from fumaroles and the plume), allowing the S magmatic signature to be estimated asδ34S = 1.0 ± 1.5‰.
 3. Redox conditions of ΔNNO ≥ 1 must exist in the shallow portion of the Etnean plumbing system, being more oxidizing than previously thought. Such conditions are also able to explain the dominant sulfur degassing in the form of SO2 measured in the plume.
 This work demonstrates how the investigation of sulfur isotopic composition in both the melt and gaseous phases may indirectly allow to constrain the redox conditions of a magmatic system.
 We wish to thank Timothy H. E. Heaton for his support during analysis of sulfur isotopes. We are also grateful to the Editor James Tyburczy and to Franco Tassi for their helpful suggestions that improved the manuscript.