Impulsive Supply of Volatile-Rich Magmas in the Shallow Plumbing System of Mt . Etna Volcano

Magma dynamics at Mt. Etna volcano are frequently recognized as the result of complex crystallization regimes that, at shallow crustal levels, unexpectedly change from H2O-undersaturated to H2O-saturated conditions, due to the impulsive and irregular arrival of volatile-rich magmas from mantle depths. On this basis, we have performed hydrous crystallization experiments for a quantitative understanding of the role of H2O in the differentiation of deep-seated trachybasaltic magmas at the key pressure of the Moho transition zone. For H2O = 2.1–3.2 wt %, the original trachybasaltic composition shifts towards phonotephritic magmas never erupted during the entire volcanic activity of Mt. Etna. Conversely, for H2O = 3.8–8.2 wt %, the obtained trachybasalts and basaltic trachyandesites reproduce most of the pre-historic and historic eruptions. The comparison with previous low pressure experimental data and natural compositions from Mt. Etna provides explanation for (1) the abundant release of H2O throughout the plumbing system of the volcano during impulsive ascent of deep-seated magmas; (2) the upward acceleration of magmas feeding gas-dominated, sustained explosive eruptions; (3) the physicochemical changes of gas-fluxed magmas ponding at shallow crustal levels; and (4) the huge gas emissions measured at the summit craters and flank vents which result in a persistent volcanic gas plume.


Introduction
The plumbing system of Mt.Etna volcano (Sicily, Italy) has a multifaceted geometry, variable in space and time and consisting of storage zones at different depths, where more or less primitive magmas containing variable H 2 O contents undergo degassing, fractional crystallization and mixing processes (cf.[1]).For example, the explosive activity of the volcano is ascribed to the impulsive upward migration of gas-rich magmas and/or fluxes of abundant volatiles from the depths [2][3][4].Evidence of pulsating volatile flushing is also provided by the great volume of gases released from the summit craters and flanks of the volcano that clashes with the relatively small amount of erupted products [5].The release of volatiles has drastic effects on the rheology of Etnean magmas and, consequently, the internal dynamics of the volcano [6].Moreover, the large quantities of gas released may change the geochemical and isotopic compositions of near-liquidus lavas erupted at the vents and flowing onto the surface [7].

Starting Materials and Methods
The starting materials used in this study are two trachybasalts (i.e., HE and ME hawaiitic products) from Mt. Etna showing virtually identical compositions: on anhydrous basis, HE and ME have 47.78 and 48.27 wt % SiO 2 , and 5.87 and 5.79 wt % Na 2 O + K 2 O, respectively (Table 1).HE is a lava flow erupted from the western flank of the Valle del Bove during the 1991-1993 effusive activity [22].ME is a lava flow erupted from the South-East Crater during the 2011-2013 paroxysmal activity comprising both Strombolian explosions and lava effusions [23,24].These two compositions are, indeed, the most representative of the recent activity at Mt. Etna volcano.Both experiments and analyses were conducted at the Department of Earth Sciences, Sapienza-University of Rome, Italy.The starting glass was produced by melting the powdered sample at 1350 • C and atmospheric pressure for 1 h.Some glass chips were then analyzed by scanning electron microscopy, and no crystalline phases were detected.Phase equilibrium crystallization experiments were performed at 800 MPa in a half-inch, end-loaded piston cylinder apparatus.The starting material, composed of the powdered glass and deionized H 2 O, was loaded into a Au 75 Pd 25 capsule.Variable amounts of H 2 O were added to the charges in order to obtain melt-H 2 O concentrations in the range between 2.1 and 10 wt % (Table 2), as determined by the "difference from 100" method (i.e., based on the difference to 100% of the total obtained by microprobe analyses of glasses, with accuracy of ~0.5 wt % H 2 O; [25]. According to H 2 O solubility data ([19] see discussion below) and the absence of gas bubbles into the quenched charges, the experiments were H 2 O-undersaturated, thus reproducing the crystallization conditions observed for Etnean magmas residing at depth [26].The assembly consisted of fluorite-graphite-magnesia.The charge was heated directly to the target temperature (i.e., between 1020 and 1200 • C) that was kept constant for 8 h (Table 2).The temperature was controlled by a factory calibrated thermocouple (W 95 Re 5 -W 74 Re 26 ; type D) with accuracy of ± 3 • C. The experiment was quenched through a cooling rate of approximately 150 • C/s and then the run product was mounted in an epoxy disk to expose a polished surface.
All the runs were self-buffered; we attempted to estimate f O 2 through the equation of [27] using the liquid Fe 3+ /Fe tot mole ratios from olivine-melt Fe-Mg exchange calculated according to [28].This procedure yielded f O 2 values of −0.3 to +0.7 NNO, in agreement with those estimated for similar furnace assemblages [29,30].
The chemical analyses were performed at the CNR-Istituto di Geologia Ambientale e Geoingegneria (Rome, Italy) using a Cameca SX50 electron microprobe (CAMECA, Gennevilliers, France) equipped with five wavelength-dispersive spectrometers using 15 kV accelerating voltage, 15 nA beam current for minerals and 2 nA for glass to obviate the alkali loss during the analyses, particularly for hydrous glasses.Moreover, sodium and potassium were analyzed first to prevent alkali migration effects.The beam diameter was of 10 µm, and the counting time was 20 s.The following standards were used: wollastonite (Si and Ca), corundum (Al), diopside (Mg), andradite (Fe), rutile (Ti), orthoclase (K), jadeite (Na), apatite (P) and metal (Mn).
Figure 1 shows that the phase assemblage is characterized by the ubiquitous occurrence of clinopyroxene and titanomagnetite at near-liquidus conditions, followed by the crystallization of plagioclase and olivine at lower temperatures and/or H 2 O contents.In highly crystallized charges, plagioclase is the most abundant mineral phase but its stability field progressively decreases with increasing H 2 O.The crystallization of clinopyroxene and plagioclase increases by 50 vol % with decreasing temperature (Table 2).In particular, the plagioclase/clinopyroxene ratio is 0.7-1.2 at T ≥ 1125 • C, but abruptly increases to 1.5-2.2 at T ≤ 1090 • C due to the predominant crystallization of plagioclase (Table 2).It is interesting to note that olivine crystallization occurs only at T ≤ 1150 • C and is subordinated to that of other mineral phases.As the temperature decreases, the near-liquidus surface is gained by increasing amounts of H2O that lower the degree of crystallization and buffer the melt composition to trachybasalt.Cpx, clinopyroxene.Timt, titanomagnetite.Plg, plagioclase.Ol, olivine.
The ulvospinel (Usp) content in titanomagnetite progressively increases along an almost linear trend resulting from the combined effect of both temperature and H2O (Figure 3c).This compositional Overall, the anorthite (An) molecule in plagioclase shows a rough decrease (from An 75 to An 63 ) with decreasing temperature (Figure 3b).However, An in plagioclase is also influenced by the initial H 2 O content, according to the dependence of CaO-NaO exchange on hydroxyl groups dissolved in the melt [34].
The ulvospinel (Usp) content in titanomagnetite progressively increases along an almost linear trend resulting from the combined effect of both temperature and H 2 O (Figure 3c).This compositional change is also associated with lower Al and Mg concentrations into the titanomagnetite crystal lattice, due to their preferential incorporation in clinopyroxene and plagioclase (Table 3).
At H 2 O = 2.3-3.8wt %, the amount of forsterite (Fo) in olivine decreases from 74 to 65 with decreasing temperature, whereas scattered Fo values are observed at the H 2 O = 2.1 and 5 (Figure 3d).As documented for plagioclase, hydroxyl groups dissolved in the melt preferentially complex with Mg 2+ rather than Fe 2+ , thus reducing the activity of MgO relative to FeO in olivine, and viceversa [34].

Achievement of Equilibrium
In order to test the achievement of equilibrium crystallization in the experimental charges, different textural and compositional features have been considered: (1) the euhedral shape of crystals, (2) the homogeneous distribution of mineral phases in the glass, (3) the lack of quench crystals, (4) the smooth variation of phase compositions with experimental conditions, (5) the small sum of residuals (generally ≤0.5;Table 2) from mass-balance calculations, and (6) crystal-melt exchange coefficients (Kd) for clinopyroxene and plagioclase comparable to those from previous equilibrium studies.Concerning the clinopyroxene phase, we have used the temperature-sensitive cpx-melt Kd Fe-Mg model derived by Putirka [35] and based on deviations in observed and calculated Fe-Mg cation partitioning between crystals and coexisting melts.cpx-melt Kd Fe-Mg values of clinopyroxene-melt pairs from this study, calculated with the total Fe as Fe 2+ according to Putirka et al. [36], vary between 0.26 and 0.27 (Table 2), in agreement with the equilibrium range of 0.27 ± 0.03 reported in literature (e.g., [36]).As a further test for clinopyroxene, we have also adopted the model of Mollo et al. [37], calibrated through the difference between diopside + hedenbergite components (∆DiHd) predicted for clinopyroxene via regression analysis of clinopyroxene-melt pairs in equilibrium conditions, and those measured in the analyzed crystals (Table 2).∆DiHd values derived for our experimental compositions are remarkably low (0.03-0.08) and close to zero, as expected for thermodynamically equilibrated phases (e.g., [38]).
The equilibrium crystallization of plagioclase has been tested through the Ab-An (albite-anorthite) exchange reaction proposed by Putirka [35].According to this model, the equilibrium constant is constrained within two temperature-dependent intervals of plg-melt Kd Ab-An = 0.10 ± 0.05 at T < 1050 • C and plg-melt Kd Ab-An = 0.25 ± 0.11 at T > 1050 • C. Our estimates show that plg-melt Kd Ab-An yields values between 0.15 and 0.21, suggesting equilibrium crystallization for experimental plagioclase crystals formed at T > 1050 • C.

H 2 O-Undersaturated Versus H 2 O-Saturated Crystallization Regimes
Previous experiments conducted at 200-400 MPa on Etnean magmas have demonstrated that olivine does not crystallize under H 2 O-undersaturated conditions [12,17,39], whereas at P H2O = P total its stability field expands at the expense of other minerals, and olivine appears as liquidus phase [16].A similar result has been also documented by the experimental study of Di Carlo et al. [40] for a K-basalt from Stromboli, where clinopyroxene dominates the phase assemblage at 400 and 300 MPa.Clinopyroxene and olivine co-saturate the melt at 200 MPa and olivine becomes the liquidus phase at 100 MPa and H 2 O > 3.5 wt % [40].Our experiments confirm that olivine is a late crystallizing phase from H 2 O-undersaturated magmas equilibrated at high pressure.This agrees with the petrological observation of Tanguy et al. [41] that clinopyroxene oversteps olivine crystallization at the Moho depth (28 km, ~800 MPa; [18]), for most of the trachybasaltic lava flows erupted at Mt. Etna.Moreover, thermobarometric estimates demonstrate that the initial stages of crystallization of Etnean trachybasalts take place near the Moho boundary [21], pointing to a deep feeder level of the plumbing system.We are aware of the critical role of CO 2 on the activity of H 2 O in the melt and degassing/crystallization processes at Mt. Etna volcano, especially at high pressure, where the solubility of CO 2 may be significant (e.g., up to ~0.7 wt % at 800 MPa; [19]).However, previous experiments on Etnean trachybasalts show that phase relations and compositions are weakly affected by CO 2 concentration in the melt, successfully reproducing the textures and compositions of natural products under CO 2 -absent experimental conditions.For example, Vetere et al. [17] have performed experiments on Etnean trachybasalts equilibrated at 400 MPa, 1100-1150 • C and NNO + 3.7 buffering conditions, as well as in presence of volatile components of 1-3 wt % H 2 O and 0-0.3 wt % CO 2 .Results from these experiments indicate that the role of CO 2 in controlling the final composition of magmas is negligible with respect to that played by T and H 2 O at degrees of crystal fractionation in the range of 6-40 wt %.More specifically, in terms of major and rare earth elements, the Etnean compositions are well reproduced in laboratory irrespective of the C-O-H species in the melt phase.Logically, an increasing concentration of CO 2 in the melt reduces the solubility of H 2 O, thus modulating the degree of crystal fractionation, as it is obtained by lowering the amount of H 2 O added in the experimental charge.
In Figure 4, the compositions of pre-historic, 1963-1995, and 2001-2006 lava flows at Mt. Etna are compared with those experimentally derived at high (P = 800 MPa), moderate (P = 200-400 MPa) and low (P = 27-120 MPa) pressures (Table S1, Supplementary Materials).In H 2 O-undersaturated experiments conducted at both high and moderate pressures, plagioclase crystallization is favored at low H 2 O content, leading to a relatively high plagioclase/clinopyroxene ratio (Plg/Cpx = 0.7-2.2).As a consequence, the CaO/K 2 O ratio in the melt decreases significantly at the early stage of crystallization, reproducing the bulk rock compositions of 2001-2006 eruptions, as well as part of the evolutionary trend of the less differentiated historic eruptions.In contrast, the high H 2 O content in the low pressure experiments suppresses the crystallization of plagioclase along most of the magma differentiation path.Plagioclase starts to crystallize at low temperature (1075 • C) and the plagioclase/clinopyroxene ratio remains relatively low (Plg/Cpx = 0.3-0.9).Thus, the experimental melts exhibit high CaO/K 2 O ratios that capture the bulk rock compositions of 1763-1995 eruptions.
According to Figure 4, it is apparent that the differentiation of Etnean magmas is controlled by the stability of plagioclase that, in turn, depends on the amount of H 2 O dissolved in the melt.At very shallow crustal levels, the core-to-rim compositional variation of plagioclase phenocrysts is driven by the exsolution of ~3 wt % H 2 O during magma decompression from 85 to 5 MPa [12].Conversely, at P > 100 MPa, disequilibrium textures of plagioclase testify to mixing induced by recharge of volatile-rich magmas or injection by volatile flushing into the system [9,24].This applies especially to the onset of lava fountain eruptions that are triggered by the irregular arrival of primitive, volatile-rich magma batches that mixed with more evolved, H 2 O-undersaturated trachybasalts residing in the upper conduit region of the volcano [1,23].
In order to explain the most differentiated trachyandesitic lavas erupted during the pre-historic activity of Mt.Etna, we have modelled the evolution of the HE trachybasalt using the MELTS algorithm of Ghiorso and Sack [42] assuming a fractional crystallization process.The best fitting differentiation path is obtained at 1040-1160 • C, 150 MPa, 1.5 wt % H 2 O and NNO buffer, when the system is H 2 O-undersaturated, but the H 2 O concentration increases from 1.5 to 3.2 wt % during crystallization (Table S2, Supplementary Materials).Under such circumstances, the plagioclase/clinopyroxene ratio is intermediate (Plg/Cpx = 0.3-1.8) to that experimentally derived at low and high pressure conditions (Figure 4), thus reproducing the bulk rock trachyandesitic compositions of natural products.This finding confirms the differentiation model proposed by [41], where pre-historic trachybasaltic magmas underwent occasional pulses of low-pressure plagioclase fractionation causing the eruption of more evolved trachyandesitic lava flows.Importantly, high-pressure experiments from this study point out that, at H 2 O = 2.1-3.2wt %, the degree of crystallization is generally very high (~40-50 vol %), so that deep-seated trachybasaltic magmas shift towards phonotephritic compositions never erupted during the entire volcanic activity of Mt.Etna.S2, Supplementary Materials).Under such circumstances, the plagioclase/clinopyroxene ratio is intermediate (Plg/Cpx = 0.3-1.8) to that experimentally derived at low and high pressure conditions (Figure 4), thus reproducing the bulk rock trachyandesitic compositions of natural products.This finding confirms the differentiation model proposed by [41], where pre-historic trachybasaltic magmas underwent occasional pulses of low-pressure plagioclase fractionation causing the eruption of more evolved trachyandesitic lava flows.Importantly, highpressure experiments from this study point out that, at H2O = 2.1-3.2wt %, the degree of crystallization is generally very high (~40-50 vol %), so that deep-seated trachybasaltic magmas shift towards phonotephritic compositions never erupted during the entire volcanic activity of Mt.Etna.

H 2 O Release upon Magma Ascent during 2001 and 2006 Eruptions
Through the thermodynamically based equation of state proposed by Duan [19], we have estimated the solubility of H 2 O and CO 2 in Etnean magmas.One of the advantages of this model consists of its applicability to high pressure (>500 MPa) and high temperature conditions that are not adequately reproduced by previous equations (see [20] for a review).Our calculations referred to 100-800 MPa and 1200 • C using the HE trachybasalt as input composition (Table 1); notably, replicated calculations at 1100-1150 • C with more or less differentiated trachybasalts (MgO = 4-6 wt % as for 2001-2012 Etnean lavas; [12]) provided changes in estimate (6.5%) within the accuracy of the model for the prediction of H 2 O (7.2%) and CO 2 (9.7%).Results are plotted in Figure 5a together with the volatile contents derived by melt inclusion data from 2001 and 2006 lava flows [3,14,43].Generally, the molar fraction of H 2 O in the gas phase ranges from 20 to 40 mol % as documented by Collins et al. [3] and, at ~400 MPa, the maximum concentrations of H 2 O and CO 2 are 3.5 wt % and 0.3 wt %, respectively (Figure 5a).The amount of volatiles dissolved in the 2001 magma is also higher than that measured for the 2006 eruption.Indeed, the 2001 lava flow was fed by a volatile-rich primitive basalt rising along a closed-system degassing path [14,43], whereas the 2006 eruption involved the slow effusion of a gas-poor magma ponding within the volcanic edifice [3,44].An important outcome from our calculations is that, at 800 MPa, the solubility of H 2 O can be very high, ranging from 6 to 8.5 wt % (Figure 5a).Therefore, trachybasaltic magmas stored at the Moho depth may be source of sustained H 2 O supply during upward migration throughout the plumbing system of the volcano.
primitive basalt rising along a closed-system degassing path [14,43], whereas the 2006 eruption involved the slow effusion of a gas-poor magma ponding within the volcanic edifice [3,44].An important outcome from our calculations is that, at 800 MPa, the solubility of H2O can be very high, ranging from 6 to 8.5 wt % (Figure 5a).Therefore, trachybasaltic magmas stored at the Moho depth may be source of sustained H2O supply during upward migration throughout the plumbing system of the volcano.Using a specific algorithm based on the equilibrium partitioning between mineral and melt, Armienti et al. [21] estimated the near-liquidus temperature of Etnean trachybasaltic magmas at the early stage of clinopyroxene crystallization.At ~800 MPa, the authors found that clinopyroxene saturation occurred at 1120-1200 • C and 1080-1150 • C for the 2001 and 2006 composition, respectively (Figure 5b).Looking at our experimental data, clinopyroxene is in equilibrium with a trachybasaltic magma when the degree of crystallization is ≤13 vol % (Table 2), otherwise the melt composition would readily evolve towards either basaltic trachyandesite or phonotephrite as a function of H 2 O (Figure 2).The trachybasaltic melts experimentally-derived in this study confirm that the temperatures estimated by Armienti et al. [21] are close to the near-liquidus region of the 2001 and 2006 eruptions.Figure 1 shows that, as the temperature decreases, the near-liquidus surface is controlled by increasing amounts of H 2 O.According to this experimental curve, we found that the original H 2 O concentrations in equilibrium with 2001 and 2006 magmas were 3-6 wt %.Coherently, these H 2 O contents closely match with the hygrometric estimates of 3-6 and 5.5-7 wt % for 2001 and 2006 eruptions, respectively, reported by Armienti et al. [21] (Figure 5b).With respect to the degassed melt inclusions found at the storage level of ~200 MPa (see also [1]), our experimental data provide quantification for the overall release of H 2 O (1.5-4 wt %) that could be expected during impulsive ascent of deep-seated trachybasaltic magmas at Mt. Etna [4,45,46].A migration of volatiles (particularly H 2 O) transported as supercritical fluids throughout the Etnean plumbing system is also consistent with the early "volatile-induced differentiation model" proposed by Ferlito and Lanzafame [8] and further reappraised by Ferlito [47] to explain the potassium enrichment observed in post-1971 lava flows and the transition from a low-density H 2 O-melt solution to a high-density melt phase erupted concomitant to a sustained gas flux.Therefore, the huge flux of magmatic gases from depth provides explanation for (1) the low volume of magma erupted in recent years with respect to the great amount of gases measured in the volcanic plume [5,47], (2) the compositions of unusually explosive eruptions fed by primitive, volatile-rich magmas [48], and (3) the textural and compositional variations of plagioclases related to fluid-induced dissolution reactions [9].In this context, Figure 6 shows that, at 800 MPa and 1070-1090 • C, the basaltic trachyandesitic magma is saturated with clinopyroxene when the amount of H 2 O is 5.2-8 wt %.Conversely, at 200-400 MPa and 1050-1070 • C, the H 2 O content decreases to 2.5-4 wt %, implying that at least 50% of H 2 O is released as a free fluid phase.Both melt inclusion data [3,43] and clinopyroxene-based hygrometric estimates [39,49] confirm that H 2 O degassing starts at ~400 MPa, with most of the H 2 O loss occurring at P < 100 MPa.
melt inclusions found at the storage level of ~200 MPa (see also [1]), our experimental data provide quantification for the overall release of H2O (1.5-4 wt %) that could be expected during impulsive ascent of deep-seated trachybasaltic magmas at Mt. Etna [4,45,46].A migration of volatiles (particularly H2O) transported as supercritical fluids throughout the Etnean plumbing system is also consistent with the early "volatile-induced differentiation model" proposed by Ferlito and Lanzafame [8] and further reappraised by Ferlito [47] to explain the potassium enrichment observed in post-1971 lava flows and the transition from a low-density H2O-melt solution to a high-density melt phase erupted concomitant to a sustained gas flux.Therefore, the huge flux of magmatic gases from depth provides explanation for (1) the low volume of magma erupted in recent years with respect to the great amount of gases measured in the volcanic plume [5,47], (2) the compositions of unusually explosive eruptions fed by primitive, volatile-rich magmas [48], and (3) the textural and compositional variations of plagioclases related to fluid-induced dissolution reactions [9].In this context, Figure 6 shows that, at 800 MPa and 1070-1090 °C, the basaltic trachyandesitic magma is saturated with clinopyroxene when the amount of H2O is 5.2-8 wt %.Conversely, at 200-400 MPa and 1050-1070 °C, the H2O content decreases to 2.5-4 wt %, implying that at least 50% of H2O is released as a free fluid phase.Both melt inclusion data [3,43] and clinopyroxene-based hygrometric estimates [39,49] confirm that H2O degassing starts at ~400 MPa, with most of the H2O loss occurring at P < 100 MPa.

Conclusions
High-pressure crystallization experiments carried out on a trachybasaltic composition, under H 2 O-undersaturated and H 2 O-saturated conditions, shed new light on the role of H 2 O on the high-pressure crystallization path of magma at Mt. Etna.We found that the crystallization sequence, phase stability, textural and compositional evolution of natural minerals are directly correlated to the transition from H 2 O-undersaturated to H 2 O-saturated crystallization regimes.Olivine is a late crystallizing phase from H 2 O-undersaturated magmas equilibrated at high pressure.In contrast, the clinopyroxene saturation surface is encountered at 800 MPa, when the solubility of H 2 O can be very high, ranging from 6 to 8.5 wt %.During upward migration throughout the plumbing system of the volcano, the overall release of H 2 O accompanying clinopyroxene crystallization is constrained between 1.5 and 4 wt %.Therefore, the impulsive ascent of deep-seated trachybasaltic magmas may release a great amount of H 2 O.Most of the excess of H 2 O may contribute to the gas emissions from the summit craters and volcano flanks feeding the persistent gas plume at Mt. Etna.However, high-pressure experiments conducted at H 2 O-undersaturated conditions cannot account for the distinctly non-equilibrium processes related to the large gas flux that may control magma compositions and trigger eruptions at Mt. Etna volcano.Therefore, further laboratory investigations on decompression and volatile degassing mechanisms may be of help to elucidate magma dynamics related to intense supply of a gas phase (mostly H 2 O) in the uppermost section of the feeding system.

17 Figure 1 .
Figure 1.Temperature vs. H2O diagram showing the stability of mineral phases in our experimental charges.As the temperature decreases, the near-liquidus surface is gained by increasing amounts of H2O that lower the degree of crystallization and buffer the melt composition to trachybasalt.Cpx, clinopyroxene.Timt, titanomagnetite.Plg, plagioclase.Ol, olivine.

Figure 1 .
Figure 1.Temperature vs. H 2 O diagram showing the stability of mineral phases in our experimental charges.As the temperature decreases, the near-liquidus surface is gained by increasing amounts of H 2 O that lower the degree of crystallization and buffer the melt composition to trachybasalt.Cpx, clinopyroxene.Timt, titanomagnetite.Plg, plagioclase.Ol, olivine.

Figure 2 .
Figure 2. Total alkali vs. silica diagram showing the composition of residual melts from our experiments.Two distinct trends are depicted as a function of the H2O content.For H2O = 2.1-3.2wt %, the original trachybasaltic liquid evolves towards basaltic trachyandesitic and phonotephritic compositions.For H2O = 3.8-8.2wt %, the starting melt differentiates to basaltic trachyandesitic to trachyandesitic to tephriphonolitic compositions.

Figure 2 .
Figure 2. Total alkali vs. silica diagram showing the composition of residual melts from our experiments.Two distinct trends are depicted as a function of the H 2 O content.For H 2 O = 2.1-3.2wt %, the original trachybasaltic liquid evolves towards basaltic trachyandesitic and phonotephritic compositions.For H 2 O = 3.8-8.2wt %, the starting melt differentiates to basaltic trachyandesitic to trachyandesitic to tephriphonolitic compositions.

Figure 3 .
Figure 3. Mineral compositional variations as a function of temperature and melt-H2O content.At H2O = 2.1-3.2wt %, the ferri-calcium Tschermak molecule (CaFeTs) slightly decreases with decreasing temperature, whereas at H2O = 3.8-8.2wt %, both Al iv and Fe 3+ are preferentially incorporated into clinopyroxene crystal lattice, leading to higher CaFeTs contents (a).The anorthite (An) molecule in plagioclase generally decreases with decreasing temperature with the exception of compositions derived upon the effect of H2O = 2.1 and 5 wt % (b).The ulvospinel (Usp) content in titanomagnetite progressively increases along an almost linear trend resulting from the combined effect of both temperature and H2O (c).At H2O = 2.3-3.8wt %, the amount of forsterite (Fo) in olivine decreases with decreasing temperature, whereas scattered Fo values are observed at H2O = 2.1 and 5 wt % (d).

Figure 3 .
Figure 3. Mineral compositional variations as a function of temperature and melt-H 2 O content.At H 2 O = 2.1-3.2wt %, the ferri-calcium Tschermak molecule (CaFeTs) slightly decreases with decreasing temperature, whereas at H 2 O = 3.8-8.2wt %, both Al iv and Fe 3+ are preferentially incorporated into clinopyroxene crystal lattice, leading to higher CaFeTs contents (a).The anorthite (An) molecule in plagioclase generally decreases with decreasing temperature with the exception of compositions derived upon the effect of H 2 O = 2.1 and 5 wt % (b).The ulvospinel (Usp) content in titanomagnetite progressively increases along an almost linear trend resulting from the combined effect of both temperature and H 2 O (c).At H 2 O = 2.3-3.8wt %, the amount of forsterite (Fo) in olivine decreases with decreasing temperature, whereas scattered Fo values are observed at H 2 O = 2.1 and 5 wt % (d).

Figure 5 .
Figure 5. H2O vs. CO2 solubility diagram derived through the model of Duan [19] (a).Calculations were referred to 100-800 MPa and 1200 °C using the HE trachybasalt as input composition.The volatile contents derived by melt inclusion data from 2001 and 2006 lava flows [3,14,43] are also plotted in figure.P-T-H2O diagram derived for the stability of clinopyroxene phenocrysts from 2001 and 2006 lava flows (b).The saturation pressure (~800 MPa) and temperature (1120-1200 °C and 1080-1150 °C for the 2001 and 2006 clinopyroxenes, respectively) were estimated by Armienti et al. [21].Our experimental data, indicate that the original amounts of H2O in equilibrium with 2001 and 2006 magmas were 3-6 and 5.5-7 wt %, respectively.Error bars come from thermometric and hygrometric estimates of Armienti et al. [21].

Figure 5 .
Figure 5. H 2 O vs. CO 2 solubility diagram derived through the model of Duan [19] (a).Calculations were referred to 100-800 MPa and 1200 • C using the HE trachybasalt as input composition.The volatile contents derived by melt inclusion data from 2001 and 2006 lava flows [3,14,43] are also plotted in figure.P-T-H 2 O diagram derived for the stability of clinopyroxene phenocrysts from 2001 and 2006 lava flows (b).The saturation pressure (~800 MPa) and temperature (1120-1200 • C and 1080-1150 • C for the 2001 and 2006 clinopyroxenes, respectively) were estimated by Armienti et al. [21].Our experimental data, indicate that the original amounts of H 2 O in equilibrium with 2001 and 2006 magmas were 3-6 and 5.5-7 wt %, respectively.Error bars come from thermometric and hygrometric estimates of Armienti et al. [21].

Figure 6 .
Figure 6.Temperature vs. H2O diagram derived using data from hydrous crystallization experiments conducted at high (800 MPa from this study) and moderate (200-400 MPa from literature) pressures.Considering the natural P-T constrains of 2011-2013 lava fountains, the amount of H2O in equilibrium with clinopyroxene phenocrysts has been reconstructed.At 800 MPa and 1070-1090 °C, the basaltic trachyandesitic magma is saturated with clinopyroxene when the amount of H2O is 5.2-8 wt %.Conversely, at 200-400 MPa and 1050-1070 °C, the H2O content decreases to 2.5-4 wt %.

Figure 6 .
Figure 6.Temperature vs. H 2 O diagram derived using data from hydrous crystallization experiments conducted at high (800 MPa from this study) and moderate (200-400 MPa from literature) pressures.Considering the natural P-T constrains of 2011-2013 lava fountains, the amount of H 2 O in equilibrium with clinopyroxene phenocrysts has been reconstructed.At 800 MPa and 1070-1090 • C, the basaltic trachyandesitic magma is saturated with clinopyroxene when the amount of H 2 O is 5.2-8 wt %.Conversely, at 200-400 MPa and 1050-1070 • C, the H 2 O content decreases to 2.5-4 wt %.

Table 1 .
Composition of HE and ME starting materials.SD is the standard deviation.
Notes: FeO tot is the total iron given as FeO.

Table 2 .
Experimental conditions, mass balance calculations, and tests for equilibrium crystallization conditions.