Deciphering the evolution of the internal dynamics of magmatic plumbing systems and identifying the key parameters that drive such dynamics are major goals of modern volcanology. Here we present a novel petrological approach that combines kinetic modelling of the diffusive relaxation of chemical zoning patterns in olivine crystals with thermodynamic modelling (MELTS) to constrain the nature and evolution of the plumbing system of Mt. Etna and the processes governing its internal dynamics. We investigated the compositional and temporal record preserved in 180 olivine crystals that were erupted between 1991 and 2008. Detailed systemization of the information stored in the sequential zoning record of the olivines reveals the existence of at least five compositionally different magmatic environments (MEs), characterized by different olivine compositions: M₀ (Fo79–83), M₁ (Fo75–78), M₂ (Fo70–72), M₃ (Fo65–69) and mm₁ (Fo73–75). Several routes of magma transfer connect these environments. We identified three prominent magma passageways between the environments M₀:M₁, M₂:mm₁ and M₁:M₂ that were active during the entire period of observation between 1991 and 2008. Modelling the diffusive relaxation of the olivine zoning patterns reveals that the transfer of magma along such routes can occur over fairly heterogeneous timescales ranging from days to 2 years. Although some of the passageways have been sporadically active in the months and sometimes years before an eruption, the magma migration activity increases clearly in the weeks and days prior to an eruptive event. In this context, major transfer routes such as M₂:mm₁ might represent temporary passageways that are activated only shortly before eruptive events. A forward modelling approach was developed using thermodynamic calculations with the MELTS software to identify the key intensive variables associated with the different magmatic environments. In this approach the observed populations of mineral compositions (e.g. Fo79–83), rather than single compositions, are associated with thermodynamic parameters [pressure, temperature, water content, oxygen fugacity (fO₂) and bulk composition of the melt] to identify the most plausible set corresponding to each ME. We found that temperature, water content and possibly oxidation state are the main distinguishing features of the different magmatic environments. The most primitive olivine population M₀ (Fo79–83) and some of its associated clinopyroxenes formed at high melt water contents (3·5–5·2 wt %), at fO₂ conditions buffered at quartz–fayalite–magnetite (QFM) or Ni–NiO (NNO), at temperatures ≥1110°C and at pressures ranging between 1·5 and 3·0 kbar (or higher). The intermediate population M₁ (Fo75–78) can be produced over a broad spectrum of conditions, but all require similar temperatures and lower water contents (0·1–1·4 wt %). The most evolved, more Fe-rich olivines of M₂ and M₃ are products of melts with much lower water contants (0·2–1·1 wt % H₂O for M₂, <0·5 wt % H₂O for M₃), and at probably somewhat more reducing (QFM) conditions at somewhat lower temperatures (∼1080°C). The M₃ environment characterized by very low water contents and reducing conditions could be related to an enhanced flux of CO₂. Combination of the characteristics of the various magmatic environments with temporal information (residence time in different environments and timing of magma transfer between these environments) allows a dynamic model of the plumbing system beneath Mt. Etna to be constructed.

Clinopyroxene–liquid geothermobarometry is a widely used tool for estimating the conditions under which mafic magmas are stored before they erupt. However, redox variability, sector zoning and disequilibrium crystallization present major challenges to the robust estimation of magma storage conditions. Moreover, most recent studies seeking to address these challenges have focused on clinopyroxenes from alkalic systems and are thus of limited use for understanding clinopyroxenes from the tholeiitic systems that dominate global magma budgets. Here we combine observations on natural clinopyroxenes from the 2014–2015 Holuhraun lava in Iceland with observations on experimental clinopyroxenes synthesized during high-pressure, high-temperature experiments on the same lava in order to investigate clinopyroxene–liquid equilibria in tholeiitic systems and optimize of geothermobarometric strategies. Natural clinopyroxenes from the 2014–2015 Holuhraun lava are sector zoned, with {1-11} hourglass sectors being enriched in the enstatite–ferrosillite component at the expense of all other components with respect to {hk0} prism sectors. In contrast with observations on clinopyroxenes from alkalic systems, sector zoning in clinopyroxenes from the 2014–2015 Holuhraun lava is characterized by differences in Ca and Na contents as well as in Ti and Al contents. The products of crystallization experiments performed at 100–600 MPa and 1140–1220 °C on a powdered starting glass at two sets of melt H2O content–oxygen fugacity conditions (∼0·1 wt % H2O and close to the graphite-oxygen redox buffer, and 0·5–1·0 wt % H2O and approximately one and half log units above the quartz–fayalite–magnetite redox buffer) demonstrate that clinopyroxene crystals from nominally equilibrium experiments can preserve strongly disequilibrium compositions. The compositional systematics of experimental clinopyroxenes are consistent with the presence of sector zoning. Furthermore, the magnitude of compositional variability increases with decreasing melt H2O content and increasing deviations of experimental temperatures below clinopyroxene liquidus temperatures (i.e. degrees of undercooling sensu lato), indicating that kinetic processes play a key role in controlling clinopyroxene compositions, even under notionally equilibrium conditions. Few published analyses of experimental clinopyroxene crystals may thus represent truly equilibrium compositions. Stoichiometric calculations on natural and experimental clinopyroxenes show that Fe3+ is a major constituent of clinopyroxenes from tholeiitic magmas under naturally relevant oxygen fugacity conditions. They also show that Fe3+ is most likely incorporated as Ca- and Al- bearing Ca–Fe-Tschermak’s component rather than Na-bearing aegirine component at oxygen fugacities up to one and a half log units above the quartz–fayalite–magnetite buffer. Elevated oxygen fugacities are thus less likely to compromise clinopyroxene–liquid geothermobarometry than previously thought. Guided by our experimental results, we combined published descriptions of clinopyroxene–liquid equilibria with geothermobarometric equations to develop an internally consistent and widely applicable method for performing geothermobarometry on tholeiitic magmas that does not require equilibrium zones to be selected a priori. Applying this method to natural clinopyroxene crystals from the 2014–2015 Holuhraun lava that formed under low but variable degrees of undercooling (perhaps 25 °C or less) returns values in excellent agreement with those from independent methods (232 ± 86 MPa, 1161 ± 11 °C). Robust estimates of magma storage conditions can thus be obtained by performing clinopyroxene–liquid geothermobarometry on tholeiitic magmas when disequilibrium is suitably accounted for.

The 1669 eruption of Mt Etna was one of the most voluminous and devastating of its flank eruptions in historical times. Despite a large body of relevant research, knowledge of the timing and duration of magma transfer and magma recharge through the internal plumbing system preceding and during the eruption is still limited. To address that lack of knowledge, we apply a three-way integrated method, linking systems analysis of crystals, a time-integrated study of zoned olivine populations, and a forward-modelling approach using thermodynamic calculations. Analysis of 202 olivine crystals erupted during the initial (pre-March 20, i.e. SET1) and the final (post-March 20; i.e. SET2 and MtRs) stages of the eruption reveals the existence of three magmatic environments (MEs) in which the majority of the olivine cores [M₁ (= Fo_75–78)] and rims [i.e. M₅ (= Fo_51–59) and M₃ (= Fo_65–69)] formed. Application of the rhyolite-MELTS software allowed us to constrain the key intensive variables associated with these MEs. We find that temperature, water content and oxidation state vary between these MEs. Application of diffusion modelling to the zoned olivine crystals allowed us to reconstruct the timing and chronology of melt and crystal transfer prior to and during the 1669 flank eruption. We find that, following the formation of the olivine cores [M₁ (= Fo_75–78)], the reservoir M₁ was intruded by batches of more evolved, degassed and possibly aphyric M₅-type magma, commencing 1·5 years prior to eruptive activity. This is the origin of the SET1 olivine rims (i.e. Fo_51–59). In the months prior to eruption, timescale data show that recharge activity along the newly established pathway M₁–M₅ increased notably. Starting in November 1668, only a few weeks after the first intrusive episode into the M₁ reservoir, a second pulse of magma injections (M₃-type magma) occurred and a new pathway M₁–M₃ opened; this is how the SET2 olivine rims (i.e. Fo_65–69) formed. For several weeks a bifurcated transport system with two dominant magma pathways developed along M₁–M₅ and M₁–M₃ dyke injections. Accompanied by vigorous seismicity, in the days immediately before eruption the local magma transfer dynamics changed and the M₁–M₅ recharge activity slowed down, as shown by a relative lack of crystals recording shorter timescales. M₁–M₃ recharge, however, remained high and persisted following the eruption onset on March 11, during which the SET1 lavas were drained. We propose that the change of the local magma transfer dynamics might be linked to changes in the local stress field brought on during eruption. This may potentially have been due to repeated dyke injections into Etna’s shallow plumbing system disrupting the early M₁–M₅ pathway and at the same time stabilizing the M₁–M₃ route as a dominant feeder. This transfer of system feeding would reproduce the observed syn-eruptive recharge and mixing in the weeks following eruption onset, culminating in the eruption of the later SET2 lavas.