Magma dehydration controls the energy of recent eruptions at Mt. Etna volcano

Olivine‐hosted melt inclusions (MIs) from tephra of the recent 2013–2018 activity at Mt. Etna were investigated for assessing the chemical evolution of magmas and quantifying their pre‐eruptive volatile budget. Microanalyses revealed two types of MIs present in all investigated eruptions; the inclusions, particularly the less evolved ones, appear to have experienced water loss coupled with SiO2 depletion. Restoration of the original SiO2‐H2O concentrations provides consistency with the thermodynamic modelling of magma evolution. The two types of MIs developed during crystallization of olivine plus clinopyroxene between 200 and 100 MPa, where magmas also experienced CO2 flushing. Degassing processes at these levels are responsible for water depletion in the melt and diffusive water loss from inclusions. Our data suggest that initial water budget is unchanged all over the last 20 years, reflecting therefore a potential in triggering highly explosive eruptions depending on degassing dynamics under open versus closed system conditions at shallow levels.


| INTRODUC TI ON
A new eruptive cycle started in January 2011 at Mt. Etna (Southern Italy). The activity was mainly explosive, with 44 episodes of lava fountaining occurred at the New South East Crater (NSEC) during the 2011-2013 period (e.g. Behncke et al., 2014;Giuffrida & Viccaro, 2017), culminating with the most powerful paroxysmal eruptions of the last 20 years at the Voragine crater (VOR) on December 2015 and May 2016 (Cannata et al., 2018). A frequency decrease in eruptive episodes marked the post-2016 activity, shifting to more effusive behaviours (Viccaro et al., 2019). Efforts in constraining spatial and temporal evolution of magma dynamics enlightened the complex plumbing system underneath the volcano (e.g. Cannata et al., 2015Cannata et al., , 2018Viccaro et al., 2016Viccaro et al., , 2019, although few data on volatile concentrations for the post-2011 magmas are currently available (Gennaro et al., 2019). Furthermore, these data suggest rather degassed magma compositions, which are in contrast with the eruptive behaviour observed at the volcano.
We provide here a new dataset of major and trace element compositions and volatile contents (i.e. H 2 O, CO 2 , S, Cl, F) in olivine-hosted appear to have experienced water loss coupled with SiO 2 depletion. Restoration of the original SiO 2 -H 2 O concentrations provides consistency with the thermodynamic modelling of magma evolution. The two types of MIs developed during crystallization of olivine plus clinopyroxene between 200 and 100 MPa, where magmas also experienced CO 2 flushing. Degassing processes at these levels are responsible for water depletion in the melt and diffusive water loss from inclusions. Our data suggest that initial water budget is unchanged all over the last 20 years, reflecting therefore a potential in triggering highly explosive eruptions depending on degassing dynamics under open versus closed system conditions at shallow levels.

Statement of significance
An extensive dataset of major and trace elements, together with H 2 O, CO 2 , S, Cl and F, has been obtained on melt inclusions entrapped in olivine crystals from three selected recent eruptions occurred at Mt. Etna volcano. These data are novel, original and timely considering that volatile concentrations for post-2011 Mt. Etna magmas are scarce in literature up to date. Assessing of the original volatile budgets of magmas and how their volatile load changes throughout the present configuration of the plumbing system is crucial to reconstruct the degassing dynamics of magmas and to identify the causes leading to energetic versus quiet eruptions of the volcano.

| COMP OS ITI ON S OF THE OLIVINE-HOS TED MIs
Olivine crystals were hand-picked from tephra smaller than 1.5 cm and prepared for in situ microanalyses on MIs including measurements of major elements, S, Cl and F by EMPA, trace element abundances by LA-ICP-MS and determination of H 2 O and CO 2 concentrations by FTIR and Raman spectroscopy (see Supporting Information 1). MIs from the three eruptive episodes have similar major element compositions after correction for post-entrapment crystallization [PEC calculated through Petrolog3 software (Danyushevsky & Plechov, 2011) is <13%; Tables S1 and S2 in Supporting Information 2]. Considering SiO 2 , CaO and alkali elements, our data suggest the existence of two types of MIs ( Figure 1). The type 1 MIs are entrapped in Fo 79-85 olivines; they are more basic (SiO 2 42.7-45.7 wt%; Mg# 55-60), with Na 2 O + K 2 O in the range 5.1-6.2 wt% and CaO 9.9-12.5 wt%. The type 2 MIs are found in Fo 69-79 olivines; they are more evolved (SiO 2 46.5-51.2 wt%; Mg# 40-54), with Na 2 O + K 2 O in the range 6.2-8.3 wt% and CaO 6.1-9.7 wt%.
Some type 1 inclusions have anomalously low SiO 2 (~42.7 wt%), which is an odd feature for volcanic rocks erupted at Mt. Etna.  (Métrich et al., 2004;Spilliaert et al., 2006). Maximum H 2 O concentrations were found in type 1 MIs, which display higher average H 2 O contents than type 2 MIs. Nonetheless, some type 1 MIs entrapped in high-Fo olivine (>Fo 80 ), especially from the December 2018 eruption, occasionally show very low H 2 O contents (~0.47 wt%), sometimes correlated with the lowest SiO 2 contents observed in type 1 MIs. Type 1 MIs are generally characterized by higher CO 2 concentrations than type 2 MIs for all selected eruptive episodes; slight CO 2 enrichment at low H 2 O content has also been observed for some type 1 MIs ( Figure 3). Sulphur  Table S3 in Supporting Information 2), also covering wider compositional ranges. Ratios of variably incompatible trace elements (i.e. La/Sm, Ba/Sr, Zr/Nb; Figure 1; Figure S1 in Supporting Information 2) support analogous geochemical signature for MIs coming from the three eruptions, with minor differences between the two types of MIs.  (Figures 1 and 2a). Furthermore, historical products show different geochemical signature if compared to volcanic rocks emitted particularly after the 1971 benchmark (e.g. Viccaro & Zuccarello, 2017). This points out that FS melts are not the parental magmas for recent products. Kohlstedt & Mackwell, 1998;Mackwell & Kohlstedt, 1990); (2) incorporation of H + in olivine through metal-vacancy defects (Demouchy & Mackwell, 2003Kohlstedt & Mackwell, 1998). A recent experimental study on Ti-depleted calc-alkaline products erupted at Klyuchevskoy volcano revealed a coupled behaviour of H 2 O and SiO 2 during re-equilibration of MIs with the external matrix (cf.

Portnyagin et al., 2019). A concomitant increase of H 2 O and SiO 2 in
MIs was observed during MIs rehydration, whereas reversal experiment showed the process is reversible.
These findings provide a possible explanation for occurrence of strong SiO 2 depletion in MIs, although no experimental studies have been conducted on olivine crystallized from Etna alkaline melts, which means data useful for disambiguation of possible mechanisms of H + incorporation and transport in Etnean olivines are not available. Few FTIR spectra collected on olivine crystals close to inclusions display, however, the presence of a broad band located at 3,160 cm −1 and a peak at 3,220 cm −1 . These characteristic bands are related to H + in point defects associated with metal vacancies (M1 for 3,160 cm −1 and M2 for 3,220 cm −1 ; Berry et al., 2005;Portnyagin et al., 2019), whereas other bands related to Si vacancies, Ti and Fe 3+ point defects were not recognized ( Figure S2 in Supporting Information 2). The correlation between low H 2 O and SiO 2 contents, together with the characteristic spectral bands of H + in metal vacancies, suggests that loss of SiO 2 and H 2 O could be due to metal defect formation in olivine crystallizing on MI walls during dehydration.
Reconstruction of the pristine H 2 O and SiO 2 contents before water loss was made for MIs of the post-2011 activity. SiO 2 content of whole rocks was used as a pristine value to recalculate original H 2 O-SiO 2 compositions in MIs, assuming that MIs were characterized initially by the same Si-saturation index as their host rocks (Portnyagin et al., 2019). Bulk rock compositions in equilibrium with olivines with variable forsteritic content were selected from the record since 2001 ( Figure S3 and Table S4 in Supporting Information 2; Viccaro & Cristofolini, 2008), avoiding samples with distinctive features inherited by accumulation of specific mineral phases (e.g. amphibole; cf. Viccaro et al., 2006). PEC was performed on restored  (Métrich et al., 2004;Spilliaert et al., 2006).
Low entrapment pressures are also linked to low CO 2 contents measured in glasses. Recent studies demonstrated that most of CO 2 in MIs can be lost from the melt to the shrinkage bubble (Hartley et al., 2014;Moore et al., 2018;Wallace et al., 2015). In order to evaluate this effect, we restored the bulk CO 2 content of MIs by adopting the approach of Wallace et al. (2015) using: (1) VOLATILECALC (Newman & Lowenstern, 2002) to determine the saturation pressure and CO 2 mol.% in the vapour phase in equilibrium with the entrapped liquid, knowing the H 2 O-CO 2 dissolved in the melt; and (2) a modified Redlich-Kwong equation of state (Kerrick & Jacobs, 1981) to calculate the molar volume of the H 2 O-CO 2 mixture in the vapour phase.  Table S5 in Supporting Information 2) was used as starting melt for the first step, constraining the initial conditions as follows: T = 1,140°C, P = 300 MPa, fO 2 at the QFM buffer, 4.02 wt% of H 2 O and 1,754 ppm of CO 2 ; the first step ended at T = 1,110°C and P = 210 MPa. The initial melt composition for the second step was fixed starting from the last melt obtained in the first step, then constraining T = 1,110°C, P = 210 MPa, fO 2 at the QFM buffer and H 2 O at 3.20 wt% and CO 2 at 1,072 ppm to take into account the effect of CO 2 flushing at 200 MPa. The final T was fixed at 1,064°C (Calvari et al., 1994) and pressure close to surface conditions. Olivine and LA-ICPMS data acquisition. Thorough reviews provided by the Associate Editor, Helena Albert and an Anonymous Reviewer greatly helped in improving the early submitted version of the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no financial or other conflicts of interests for this work.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the Supporting Information of this article.