Introduction and geological background

Textural studies of phenocryst populations in magmatic rocks represent an important approach for petrology and volcanology, as they could open a window into the geometry of the feeding systems beneath volcanoes and their magma dynamics such as melt ascent, ponding, degassing, and/or mixing (Nakamura 1995; Ginibre et al. 2007; Kahl et al. 2013; Giacomoni et al. 2014, 2016; Neave et al. 2014; Jerram et al. 2018; Nardini et al. 2022; Petrone et al. 2022). Compositional zoning in minerals develops in response to chemical gradients and variations in the crystal growth regimes due to changes in element concentration at the crystal–melt boundary layer (Gorokhova and Melnik 2010; Zhang 2010; Mollo and Hammer 2017) and often reflects different temperature and pressure (T–P) conditions of nucleation and/or different compositions of the forming melt. Crystal zoning is preserved when the diffusion rate of the chemical components is slower than the pre-eruptive storage time (Costa et al. 2008; Kahl et al. 2011; Dohmen et al. 2017). These dynamics could be recorded by all the principal magmatic rock-forming minerals like olivine (Girona and Costa 2013; Kahl et al. 2013; Pontesilli et al. 2021; Casetta et al. 2023), clinopyroxene (Morgan et al. 2004; Petrone et al. 2018; Nardini et al. 2022), orthopyroxene (Saunders et al. 2012; Mangler et al. 2020), plagioclase (Ginibre et al. 2002; Druitt et al. 2012; Neave et al. 2013; Moschini et al. 2023), amphibole (Humphreys et al. 2006), sanidine (Ginibre et al. 2004; Morgan et al. 2006), quartz (Chamberlain et al. 2014) and zircon (Griffin et al. 2002). Clinopyroxene is commonly used for tracing the chemical and physical history of plumbing systems and the timescales of magmatic processes, owing to its extended stability field over variable pressure, temperature and melt water contents (H2O) (Putirka 2008; Masotta et al. 2020), its sensitivity to the crystallization conditions (Putirka et al. 1996) and the sluggish diffusion rate of some crucial elements such as Cr, Al, Ti and Fe–Mg (Dimanov and Sautter 2000; Cherniak and Dimanov 2010; Müller et al. 2013; Ni et al. 2014; Dohmen et al. 2017). At Stromboli volcano (South Italy), the consistent presence of zoned clinopyroxene has been fundamental for describing the periodic replenishment of a degassed shallow reservoir by mafic input remobilising the crystal mush. Moreover, the study of the frequency of clinopyroxene textures has revealed the triggering nature of the rapid mixing showed by the short storage timescale in the Post-Pizzo period (1.7–1.5 ka), the extensive mush disruption and cannibalisation in the 2003–2017 activity and the mush rejuvenated phase of the 2019 paroxysms (Di Stefano et al. 2020; Petrone et al. 2018, 2022). Similarly, at Mt. Etna, Cr-enriched and Al-depleted domains in clinopyroxene from the 1974 to 2014 eruptions record punctuated episodes of intrusion of primitive magma at depth that acted as trigger of the system (Ubide and Kamber 2018). Again, mixing dynamics between evolved and more primitive melts have been described through the composition of zoned pyroxene crystals in the Mexican volcanoes of Volcán de Colima (Hughes et al. 2021) and Popocatépetl (Mangler et al. 2022).

However, indirect evidence based on petrologic models alone is not sufficient to precisely locate the main magma ponding zones beneath active volcanoes, since the link between clinopyroxene textures/compositions and specific T–P–X portions of the plumbing system is not always straightforward. In this regard, complexes that expose remnants of the plumbing system (i.e. dykes or plutonic bodies) are exceptionally useful to unlock the volcanic–plutonic link (Bachmann et al. 2007). Evaluations on the volcanic–plutonic link were investigated with different approaches on the island of Skye (Scotland, British Tertiary Igneous Province) (Thompson 1982; Hamilton et al. 1998; Stuart et al. 2000) and the Miocene Colorado River magmatism (Metcalf 1982; Wallrich et al. 2023). In both contexts, the plutons were shown to be capable of preserving a useful record of plumbing system processes, being chemically related to the erupted products.

To further enhance our understanding of magmatic systems, studies of ancient volcano-plutonic complexes should be performed in strict comparison with the record of plumbing system dynamics beneath active volcanoes. In such a framework, one of the best open-air laboratories is the Dolomites area (Southeastern Alps, Italy) where intense magmatic activity took place during the Middle Triassic in a short time span (ca. 237–242 Ma; Storck et al. 2019). In this area, large volumes of volcanoclastic sequences, plutonic rocks and dyke swarms crop out over a 2,000 km2 area, presenting an excellent state of preservation of the original tectonic/stratigraphic relationships to the country rocks. This allows the sampling of almost continuous sequences of exposed and well-preserved effusive and intrusive products (e.g. Nardini et al. 2022). The volcanic products of the Dolomites area are mostly SiO2-saturated trachybasaltic to trachyandesite, and exhibit orogenic characteristics, as they belong to the calc-alkaline and shoshonitic series and have distinct large-ion lithophile elements (LILE) enrichment and Nb, Ta and Ti negative anomalies (Casetta et al. 2018a, b; 2021; Lustrino et al. 2019; De Min et al. 2020). The stark contrast between the geochemical composition and the concurrent extensional tectonic regime (Doglioni and Carminati 2008; Gianolla et al. 2009) led to the development of several geodynamic models (see Abbas et al. 2018 and reference therein), the most widely accepted being those invoking magma genesis from a mantle source metasomatized during the older Variscan subduction (Bonadiman et al. 1994; Lustrino et al. 2019; Casetta et al. 2021). While detailed thermobarometric studies on the emplacement and genesis of the intrusive bodies in the Dolomites area have already been carried out (Bonadiman et al. 1994; Casetta et al. 2018a, b), the volcanic counterparts have remained poorly studied, leaving open questions on the architecture of the plumbing systems. The excellent exposure and wide range of compositions of the volcanic rocks and their intrusive counterparts provide a unique opportunity to study directly the usually hidden plumbing system evaluating magma dynamics modelling and testing the applicability to active volcanoes. A first detailed textural/compositional study of clinopyroxene phenocrysts in trachybasaltic lavas from the Cima Pape volcano-plutonic body showed that primitive mafic melts repeatedly intruded a mildly evolved shallow magma reservoir (Nardini et al. 2022). This left questions about the timescales of such processes and their frequency at the regional scale. In this work, a detailed study of volcanic and subvolcanic rocks from Cima Pape, Predazzo, Mt. Monzoni and Sciliar areas (Fig. 1) was performed to unravel the main magmatic processes that fuelled the activity of the Mid-Triassic province, discussing the consistency between zoning patterns found in clinopyroxene from each different eruptive centre, and thus evaluating the similarities and differences between the individual feeding systems at a local scale. A comparison with the clinopyroxene population among clinopyroxenitic xenoliths from the Latemar area was carried out to identify any connections between the deeper and shallower portions of the plumbing systems. Thermobarometry and diffusion chronometry models were used to constrain the intensive parameters (T, P), the compositional evolution of the magmas, the geometry and dynamics of the feeding systems and the timescales of the magmatic processes. Besides exploring in detail the features of the Mid-Triassic magmatism in the “volcanological laboratory” of the Dolomites area, our model can be used for comparison with active volcanoes, improving our ability to link mineral textures and compositions in volcanic rocks with architecture and dynamics of the plumbing system.

Fig. 1
figure 1

a Tectonic map of the units in the eastern portion of the Alps. LO Ligurian ophiolites, AM deformed Adriatic margin, AD Adriatic microplate, SA Southern Alps, DI Dinarides, SM Southern margin of Meliata, HB Eoalpine high-pressure belt, TW Tauern tectonic Window, EW Engadine tectonic Window, OTW Ossola–Tessin tectonic window, EA Eastern Austroalpine, H Helvetic domain, M Molasse foredeep. b The Mid-Triassic magmatic occurrences in the Southern Alps domain are evidenced in black. Simplified geological map of the centres considered in this study (Modified from Casetta et al. 2021)

Full size image

Samples and methods

Among the Mid-Triassic volcanites of the Dolomites area, we selected representative samples of fresh lavas and dykes preserving evidence of clinopyroxene compositional and textural variety. The analysed samples are from the main volcano-plutonic complexes, i.e. Predazzo, Mt. Monzoni, Cima Pape and Sciliar (Fig. 1). In addition, we analysed also representative clinopyroxenitic xenoliths entrained in trachybasaltic to lamprophyric dykes cutting a lava sequence in the Latemar carbonate platform.

Whole-rock major and trace element analyses were performed with an ARL Advant-XP automated X-Ray Fluorescence (XRF) spectrometer hosted at the Department of Physics and Earth Sciences of the University of Ferrara. The entire matrix correction process, along with the measurement of intensities, was carried out in accordance with the methods of Traill and Lachance (1966). Accuracy and precision were better than 2–5% for major elements and 5–10% for trace elements. Detection limits were 0.01 wt% and 1–3 ppm for most of the major and trace element concentrations, respectively. Mineral phase major element chemistry was determined with a CAMECA SXFive FE electron microprobe equipped with one ED and five WD spectrometers hosted at the Department of Lithospheric Research of the University of Wien. The operating conditions were as follows: 15 kV accelerating voltage, 20 nA beam current, and 20 s counting time on peak position. For glass (melt inclusions), a 5 µm defocused beam and 10 s counting time on peak position for Na and K were used. Natural and synthetic standards were used for calibration, and PAP corrections were applied to the intensity data (Pouchou and Pichoir 1991).

Back-scattered electron (BSE) images for the diffusion chronometry calculations were acquired with the Jeol IT500 FEG-SEM and the FEI QUANTA 650 FEG-SEM, at the Imaging and Analysis Centre (IAC) at the Natural History Museum in London (NHM) operating at 15–20 and 15 keV electron current, and 10–11 and 15 mm WD (working distance), respectively. Under this operating condition, the resolution is better than 50 nm, which is four times smaller than a normal pixel size of 0.2 μm (Petrone et al. 2022).

Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) element mapping of clinopyroxene was undertaken in the IAC at the NHM using a Teledyne Iridia 193 nm system coupled to an Agilent 8900 ICP-MS. The samples were ablated in a helium atmosphere at a rate of approximately 0.6 l min−1. The resulting aerosol was mixed with argon (~ 0.9 l min−1) and introduced to the interface of the ICP-MS using the aerosol rapid introduction system (ARIS). The instrument was tuned for minimum aerosol dispersion during mapping experiments by optimizing the full width of a single laser pulse at 10% of the maximum peak intensity (FW0.1 M) to less than 5 ms, using a live signal monitor (LAMonitor) in the HDIP software. This optimization improved signal sensitivity and reduced imaging time. Compositional maps of the laser-ablated samples were obtained by scanning adjacent lines while moving the stage at a constant speed. The mapping experiments employed a 5 × 5 μm square beam with a fluence of 3.5 J cm−2, a repetition rate of 400 Hz and a scan speed of 167.67 μm s−1, corresponding to 12 laser pulses per 5 μm pixel. Six isotopes were measured in a no-gas mode, with each isotope having an optimized dwell time selected as an integer of the period of the laser to ensure no aliasing artefacts. Isotopes measured were: 24 Mg, 27Al, 43Ca 47Ti (2.5 ms); 52Cr,60Ni (5 ms). Individual compositional maps took between c. 20 min and c. 1.5 h to gather. A gas blank was gathered between each successive line scan. Reference material NIST SRM 612 was measured as line scans every ~ 30 min for calibration and drift correction. The acquired images were processed and quantified using HDIP v 1.7 software. Drift correction was applied to the measured counts of each isotope using the NIST SRM 612. Pyroxene grains were isolated based on the measured counts of 43Ca using the labelled segments tool in HDIP and trace element concentrations were calculated on a pixel-by-pixel basis within HDIP, with the internal standard being the Ca concentration obtained from EMPA analysis. Glass reference materials BCR-2 g and GSD-1 g were used as secondary reference materials. For crystals showing variations in calcium concentration, each zone was isolated based on its chemical composition and the appropriate Ca concentration was used as the internal standard for each zone.

Petrography and whole-rock geochemistry

The studied volcanic products from Cima Pape, Predazzo, Mt. Monzoni and Sciliar area (Fig. 1) are lava flows, lava breccias and dykes. The volcanic products of Predazzo and Cima Pape have been previously partially studied in terms of petrography and whole-rock geochemistry (Casetta et al. 2021; Nardini et al. 2022). The new samples used in this study, are from Mt. Monzoni and Sciliar (Table 1) and they will be integrated with the existing dataset. All the studied rocks are basalts to trachybasalts in composition, with highly porphyritic textures (porphyritic index; P.I. 30–55%), Mg# [calculated as 100 × MgO/(MgO + FeO) mol%; considering an Fe2O3/FeO ratio of 0.15, in agreement with a fO2 around FMQ buffer (Kress and Carmichael 1991)] ranging from 55 to 69 and an alkali content spanning from 3.1 wt% to 6.9 wt%. The SiO2 content varies from 49.2 to 54.6 wt%, while CaO and Na2O range from 6.8 to 11.4 wt% and from 1.4 to 4.0 wt% respectively. This confirms that all the studied samples fall on the transitional-alkaline curve as the other Middle Triassic magmatic products of the area (see Lustrino et al. 2019).

Table 1 Whole-rock major and trace element composition of representative lavas and dykes from Mt. Monzoni and Sciliar (Dolomites, Southern Alps). Fe2O3 and FeO were calculated by considering a Fe2O3/ FeO ratio of 0.15, in agreement with a fO2 around FMQ buffer (Kress and Carmichael 1991). Mg# = Mg/[Mg + Fe2+] mol%; LOI = loss on ignition. P.I. = porphyricity index; Bas = basalt; Bas tr-and = basaltic trachyandesite; Trachybas = trachybasalt
Full size table

The phenocryst population is usually dominated by clinopyroxene (35–45 vol.%) and plagioclase (10–25 vol.%), with subordinated olivine and Fe–Ti oxides (5–10 vol.%). Clinopyroxene crystals are present with euhedral prismatic habit, always well preserved, and often present as glomerocrysts or crystals that reach up to 1 cm in size. Clinopyroxene often hosts numerous oxides and melt inclusions in the core. Plagioclase phenocrysts have euhedral to subhedral habits, vary from 1 to 4 mm in size and are usually altered in sericite. When fresh, they show the classic albite polysynthetic twinning at crossed plane-polarized light. Olivine is rare and usually appears altered in iddingsite. The matrix is microcrystalline and composed of acicular microlites of clinopyroxene, plagioclase and oxides.

The clinopyroxenitic xenoliths sampled in the Latemar area (Fig. 1) have a cumulate texture representing the settling of clinopyroxene by gravity, with well-preserved diopside crystals with sizes from 0.2 to 2 mm in contact by means of triple junctions. Clinopyroxene sometimes hosts small, resorbed olivine in the core (⁓60 μm). Some of the samples are infiltrated by late calcite veins propagating from the host dykes.

Clinopyroxene texture and composition

Lavas and dykes

In all the studied rocks, clinopyroxene crystals are often highly zoned, irrespectively of their occurrence as phenocrysts, micro-phenocrysts or parts of aggregates. Two compositional zones occur in all the crystals: an augitic domain (Wo38-43En37-41Fs11-20) (wollastonite, enstatite, ferrosilite), light grey in BSE images, which represents the majority of the clinopyroxene composition, and a dark grey diopsidic domain (Wo42-48, En40-49, Fs5-15) (see supplementary Table S1 and Nardini et al. 2022). These two domains are compositionally well distinct with sharp boundaries. Occasionally, a hybrid composition of the two occurs at the external boundary of the diopsidic domain. The zoning patterns shown in clinopyroxene crystals are complex and composite. Normal zoning, represented by diopsidic cores surrounded by augitic mantle and rim, occurs together with concentric zoned crystals, described by the presence of one or multiple diopsidic bands within an augitic clinopyroxene. These two zoning patterns also coexist in some individuals, resulting in crystals having a diopsidic core coated by an augitic mantle which is overgrown by one or multiple diopsidic bands with the same composition of the core. Frequently, augitic zones of the crystals display oscillatory zoning (Fig. 2). Sector-zoned crystals are often combined with the other zoning patterns (Fig. 2a, d). The augitic domain has Mg# from 67 to 78 with Al2O3 between 2.2 and 7.0 wt% and Na2O and TiO2 from 0.2 to 0.6 wt% and 0.4 to 1.8 wt%, respectively, while CaO spans from 18.2 to 22.7 wt%. Cr2O3 is mostly less than 0.1 wt%, with an average composition of 0.05 wt%. The diopsidic domain has Mg# from 77 to 91, TiO2 between 0.1 and 1.1 and Cr2O3 which increases significantly up to 1.2 wt%. Al2O3 ranges from 1.2 to 6.8 wt% and CaO from 20.2 and 23.7 wt%, while Na2O is between 0.1 and 0.4 wt% (see Supplementary Figure S1). The chemistry of the zoned clinopyroxene populations is summarized in LA-ICP-MS maps (Fig. 2) where the concentrations of Mg, Ti, Cr and Ni are displayed: diopsidic portions are characterised by the depletion of Al and Ti, and enrichment of Cr which is also associated with high Ni (up to 200–250 ppm) in stark contrast with the low Ni concentration (~ 20–30 ppm) in the augitic parts (Fig. 2).

Fig. 2
figure 2

Major and trace element chemical mapping of clinopyroxene crystals from the Dolomites. BSE images (top) and quantitative chemical mapping of Mg, Ti, Cr and Ni for four representative clinopyroxene crystals: (a) Type B1, (b and c) Type B0 (mottled core and resorbed diopsidic) and (d) B2 (highly resorbed diopsidic core) from left to right. Oscillatory zoning in the augitic domains of the crystals is noticeable in Ti and Cr chemical mapping

Full size image

The TiO2 distribution in clinopyroxene from Predazzo and Mt. Monzoni centres follows two different trends, this feature being more visible in the augitic portions of the crystals due to their high TiO2 content. The clinopyroxene crystals in Predazzo and Mt. Monzoni lavas are also more enriched in TiO2 (0.8 to 1.8 wt%) than those in dykes and all crystals in lava samples from the other three complexes, where clinopyroxene has a TiO2 concentration from 0.4 to 0.6 wt% (Supplementary Figure S1). This trend is also partially visible in Al2O3 concentration. Glomerophyres have the same compositional features of phenocrysts and could have embedded, planar or point contacts. Although the majority of the crystal population is represented by homogenous augitic unzoned crystals, zoned clusters are common. In the next paragraphs, we are going to focus on the characteristics of specific zones of the clinopyroxene.

Clinopyroxene cores

Clinopyroxene cores are either augitic or diopsidic. The distribution of the two is similar between lavas and dykes, but is different between the single volcanic centres. Normal zoning in clinopyroxene with diopsidic cores are the most common in Mt. Monzoni rocks, while in the other three centres only ⁓40% of the zoned clinopyroxenes have diopsidic cores. Augitic cores are homogenous, with Mg# from 67 to 78, Al2O3 from 2.4 to 5.8 wt%, Cr2O3 from 0 to 0.1 wt% and TiO2 from 0.4 to 1.5 wt% (Fig. 3a).

Fig. 3
figure 3

Mg# and Cr2O3 profiles of every type of core in the clinopyroxene crystals from the Dolomites. a Augitic core (Type A); (b) resorbed and (c) patchy-zoned diopsidic cores (Type B); (d) diopsidic core with crystallisation of the hybrid melt (see text for details)

Full size image

Some diopsidic cores are resorbed, with embayments and undulations on their boundaries (Fig. 3b). Their Mg# is between 77 and 88, their Al2O3 and TiO2 concentrations vary from 1.3 to 6.9 wt% and from 0.2 to 1.3 wt%, respectively, and the Cr2O3 content is within a range of 0.1–0.9 wt%. Other cores show a patchy-zoned texture (Streck 2008) with the same composition of the resorbed non-patchy cores, frequently coupled with numerous melt or oxide inclusions (i.e. spotted; Nardini et al. 2022) (Fig. 3c). Sometimes the patchy-zoned cores, especially those with inclusions, have a more evolved composition, having Mg# down to 74, but always maintaining high Cr2O3 values. In these cases, composition moves to be more salitic (Wo45-47, En39-43, Fs10-14) (see Supplementary Figure S1), with a loss of Mg contents but retention of Ca and Cr contents probably due to the faster diffusion of Mg than Ca and Cr (Cherniak and Dimanov 2010; Ubide and Kamber 2018).

Oxide inclusions entrained in the cores are magnetite and ulvospinel in composition (see Supplementary Table S2). TiO2 ranges from 2.5 to 14.8 wt% and FeOT between 57.5 and 81.9 wt%, while Al2O3 and Cr2O3 have a concentration from 1.3 to 7.7 wt% and 0.05 to 11 wt%, respectively. The melt inclusions (up to 70–80 μm in size) are glassy and homogenous, their composition is trachytic with Mg# between 6.1 and 25.9, SiO2 comprised between 59.6 and 63.1 wt% and Al2O3 from 20.4 to 21.1 wt%. The alkali content ranges from 7.2 to 11.2 wt% with Na2O from 2.5 to 4.5 wt% and K2O from 4.8 to 6.8 wt%. Their MgO concentration is < 0.6 wt%. The melt inclusions, consistent with those documented at Cima Pape (Supplementary Table S3), have the same shoshonitic affinity as the magmatic samples from the Dolomites, belonging to the same liquid line of descent. Thus, these inclusions represent the melt in equilibrium with the more evolved portions (Mg# < 70) of the clinopyroxene crystals during the final stages of magma differentiation. To a lesser extent, the core is sector-zoned, normally with enrichment in Si–Mg in the hourglass sector and in Al–Ti in the prism sector (Ubide et al. 2019a; Masotta et al. 2020), but in one single case, the two domains are inverted, presenting Al–Ti enrichment in the {-111} sector as already described by Welsch et al. (2016). Sometimes, diopsidic cores show a compositional intermediate well-distinct zone before the augitic rim, with a hybrid composition (Wo43-44, En44-45, Fs11-12) between augite and diopside, as shown by the intermediate grey step in the BSE images (Fig. 3d). Some patchy-zoned crystals could be the result of amalgamation of more clinopyroxene crystals that underwent dissolution or pervasive resorption before forming the assemblage, which often takes the shape of a euhedral single crystal (i.e. mottled texture; Streck 2008; Bennett et al. 2019; Palummo et al. 2021). Usually, the aggregates with diopsidic cores are mottled, rarely preserving the original zonation of each crystal (Supplementary Figure S2).

Clinopyroxene mantles

Mantle zones of clinopyroxene crystals are either homogeneous augitic without any zonation or variably zoned. The zoning patterns in mantles are typified by the presence of one or multiple diopsidic bands. The bands have Mg# between 78 and 87 and Cr2O3 from 0.3 to 0.8 wt%. TiO2 ranges from 0.3 to 0.6 wt% in the clinopyroxene mantles of the dyke samples and is between 0.7 and 0.9 wt% in lavas. Al2O3 ranges from 2.5 to 6.0 wt%, while CaO ranges from 20.5 to 22.5 wt%. The number and thickness of the diopsidic bands varies: bands could be up to 100–200 µm wide or less than 5 µm (Fig. 4a, b). These diopsidic bands are composed of continuous, planar segments, each of which corresponds to growth on a certain crystal face and could have sharp or gentle interfaces with the augitic domain depending on how much the diffusion has smoothened the compositional gradient (Fig. 4b). In these cases, the Mg# profile is smoother than the Cr2O3 one, as the latter is slower to diffuse than the former (Fig. 4b). The diopsidic bands sometimes show, similarly to the cores, multiple compositional plateaux, with the inner part of each band being diopsidic in composition, and the outer part having a more evolved signature, hybrid between the diopside bands and the augite (Fig. 4c). The bands are present also in glomerophyres, with all crystals that always share the same bands indicating that the latter were formed after the constitution of the glomerophyre (Fig. 4d). This last feature indicates more complex zoning, due to the attachment of different crystals with different diffusion stories, as indicated by the variable thickness of the diopsidic band through the gromerophyre.

Fig. 4
figure 4

Mg# and Cr2O3 profiles of every type of diopsidic bands in the mantle of clinopyroxene crystals from the Dolomites. a Thick diopsidic band between the augitic core and rim (Type A1); (b) thin and Fe–Mg diffused multiple diopsidic bands between the augitic core and rim (Type A2); (c) double plateau diopsidic band in a Type A1 crystal due to the crystallisation of the hybrid melt (see text for details); (d) aggregate of clinopyroxene crystals sharing the same thick diopsidic band

Full size image

The augitic unzoned mantle has Mg# between 71 to 74 and low Cr2O3 concentration ranging from 0.01 to 0.08 wt%. Al2O3 concentration is from 2.9 to 4.5 wt% and CaO from 19.8 to 21.5 wt%, while TiO2 ranges from 0.6 to 0.7 in the clinopyroxene mantle of dykes and from 1.0 to 1.2 wt% in the clinopyroxene mantle of lavas.

Clinopyroxene rims

The rims, characterized by their absence of diopsidic composition, are augitic and homogenous or oscillatory zoned (see Fig. 2). Their composition is comparable to that of the augitic mantle and core, with Mg# between 71 and 76, Al2O3 from 2.2 to 4.7 wt%, CaO from 19.9 to 21.8 wt% and Cr2O3 < 0.1 wt%. As already reported for the cores and mantles, the TiO2 content describes two different trends between crystals inside lavas and dykes: in lava samples, the rim of clinopyroxene crystals has a TiO2 concentration between 0.8 and 1.4 wt%, while in dykes the TiO2 concentration shifts down to 0.4–0.7 wt%. Texturally, all the crystal rims are euhedral without dissolution or resorbed structures. Occasionally, the rim could have irregular borders representing disequilibrium crystallisation.

Texture distribution and classification

Comparing textural and compositional features, we identified different types of clinopyroxene crystals and aggregates, integrating the classification proposed by Nardini et al. (2022). Each type was mostly documented in specific samples from different areas. A summary of the main occurrences, volumetric abundances and features is reported in Table 2. Following the previous classification, Type A and Type B include all the crystals with augitic and diopsidic cores, respectively (Nardini et al. 2022). These two groups include the subtypes A1, A2, B1 and B2, distinguishing between whether a single (1) or multiple (2) diopsidic bands mantle an augitic (A) or diopsidic core (B). In the case of B1 and B2, diopsidic cores and bands are separated by augitic portions. In addition, we have included the non-zoned population (NZ, Nardini et al. 2022) in the Type A crystals as subtype A0, having augitic core, mantle and rim. Similarly, we introduced subtype B0 to describe diopsidic cores with augitic mantle and rim without any diopsidic bands. Moreover, to incorporate the sector-zoned crystals found in Predazzo and Mt. Monzoni clinopyroxene, we adapted the previous scheme by introducing a Type C clinopyroxene, with C0 and C1 subtypes describing a normal sector zoned or sector zoned with a diopsidic band, respectively. As seen in Table 2, the clinopyroxene population in Mt. Monzoni rocks does not show diopsidic bands (A1/A2, B1/B2, C1 types not found in Mt. Monzoni rocks), while in the other centres, this feature is frequent.

Table 2 Summary of clinopyroxene crystal types and abundance in the four centres considered
Full size table

Clinopyroxenitic xenoliths

In the fresh portion of the xenoliths, clinopyroxene is diopsidic in composition (Wo48-49, En46-47, Fs4-5; Mg# = 90–92, CaO ⁓24 wt%) (Fig. 5a), with TiO2, Cr2O3 and Al2O3 contents in the range 0.05–0.2 wt%, 0.15–0.4 wt% and 0.7–1.4 wt%, respectively (see Supplementary Fig. 1). Sometimes, clinopyroxene crystals enclose small (⁓ 60 μm), resorbed olivine grains with Fo85 and an NiO content up to 0.15 wt% (Fig. 5b). A few nodules are infiltrated by calcite veins propagating from the host rock (Fig. 5c). The crystals in direct contact with the veins have Wo46-48En44-45Fs8-9, with Mg# = 83–85, TiO2 from 0.14 to 0.28 wt%, Al2O3 from 1.5 to 2.3 wt% and Cr2O3 between 0.03 and 0.07 wt%. They show a normal step zoning pattern (Fig. 5c) with more evolved rims (Wo46-51, En36-41, Fs12-13; Mg# = 75–77) having TiO2 and Al2O3 concentrations of 0.4–0.7 wt% and 4.5–6.0 wt%, respectively. This scenario is magnified in the clinopyroxene crystals in contact with the host rock, having a concentration peak in TiO2 (up to 3 wt%) and Al2O3 (up to 8 wt%).

Fig. 5
figure 5

a Classification diagram of clinopyroxene (Morimoto 1988) in the clinopyroxenitc nodules; (b) photomicrographs in the transmitted plane-polarized light and BSE image of fresh clinopyroxene and olivine inclusion; (c) photomicrographs in transmitted plane-polarized light and BSE image of clinopyroxene in contact with the calcitic vein (beneath) and host rock (above). The shape and colour of the point are the same as in (a)

Full size image

Discussion

T–P–H2O conditions of the magmatic environments

Understanding the T–P conditions of formation of the mineral phases in the volcanic samples is crucial for unravelling the architecture of plumbing systems in the Dolomites. As clinopyroxene is the only well-preserved mineral phase, cpx–melt equilibria should be used to perform thermobarometric calculations (see Putirka 2008). However, the absence of glassy material with primitive compositions and the high P.I. (nearly 50%) of the effusive rocks make this approach challenging, sometimes requiring melt compositions to be modelled (Casetta et al. 2021; Nardini et al. 2022). To overcome these problems, we used the clinopyroxene-only thermometer and barometer introduced by Wang et al. (2021) and Higgins et al. (2022). The former was selected for its inclusion of the liquid H2O content in the model, overcoming the deficiency of cpx-only thermometers, which normally overestimate the temperature (Wieser et al. 2023). Here we adopted Eq. 32dH of Wang et al. (2021) (SEE of 36.8 °C) for the temperature modelling and the Higgins et al. (2022) machine learning barometer to estimate P. This latter method gives an uncertainty associated with a given individual prediction obtained from a random forest model through a composition that has passed through 300 trees. An uncertainty on an estimate can be ascribed to the interquartile range of the prediction for all 300 trees. As the thermometer requires the estimated H2O content of the equilibrium liquid, we solved the T–P–H2O path by iterative application of the hygrometer of Perinelli et al. (2016) with the above-mentioned thermobarometers.

Thermobarometry shows that diopsidic domains record a higher temperature (1056–1170 ± 21 °C) than augitic domains (1044–1118 ± 9 °C), at comparable P (295 ± 100 MPa for the augite and 254 ± 62 MPa for the diopside) for all centres (Fig. 6a). Results indicate a mean dissolved water content of 2.8 ± 0.4wt% in both melts in equilibrium with the augitic and diopsidic domains (Fig. 6b). Clinopyroxene crystals from clinopyroxenitic nodules in the Latemar area have a similar composition to the diopsidic domains in lavas and dykes and hence record similar P ranges (350 ± 53 MPa). To retrieve information on the melt responsible for the formation of the domains and corroborate our previous observation, we have applied the machine learning chemometer of Higgins et al. (2022). Calculations show that the augitic domains were in equilibrium with a trachyandesitic melt, with Mg# 45, while the diopsidic domains were in chemical equilibrium with a more primitive trachybasaltic liquid with Mg# 56. These results, combined with the textural features, corroborate the mixing model presented for the Cima Pape complex (Nardini et al. 2022) where a trachyandesitic magmatic reservoir is periodically refilled by a more primitive and hotter trachybasaltic magma.

Fig. 6
figure 6

a T–P conditions of clinopyroxene in the volcanic and dyke rocks of the four Dolomitic centres; (b) estimation of H2O content in the melt in equilibrium with augitic and diopsidic domain compared to Mg# of clinopyroxene

Full size image

Diopsidic mantles of type A1 and A2 clinopyroxene crystals with pre-existing augitic cores are expressions of the mafic inputs that periodically refilled the trachyandesitic environment. The augitic cores are interpreted to be antecrysts formed in a trachyandesitic magma different from their ultimate carrier magma, which is the result of multiple mixing events with refilling magmas upon complete homogenisation (Fig. 7a, c). Recurrent oscillatory zoning in the augitic rim domain of the clinopyroxene population (see Fig. 2a) indicates slightly different T–X conditions during the growth of the augitic domain due to convective dynamics as confirmed by the absence of oscillatory zoning in Ni (Fig. 2; Di Fiore et al. 2021). This, alongside with the large presence of Type A0 clinopyroxene in all the products, suggests a melt-dominated magmatic environment.

Fig. 7
figure 7

a Sketch of a general feeding system beneath the Mid-Triassic volcanoes in the Dolomites. The mafic input recycles diopsidic antecrysts permeating the diopsidic mush (witnessed by clinopyroxenitic nodules) in the shallower trachyandesitic magma reservoir (Type B crystals) and mantle in the augitic antecrysts (previously formed in the more evolved reservoir) with diopsidic bands (Type A1 and A2 crystals). b Examples of crystals in the Dolomites outline the main textural differences between the Mt. Monzoni zoned clinopyroxene population and the clinopyroxene found in the other three complexes. c Flowchart summarizing the main processes acting in the two magmatic environments fossilised in Type A and B clinopyroxene

Full size image

Despite the wide range of pressure estimates (see Fig. 6a), field characteristics of intrusive bodies in the area suggest that the trachyandesitic reservoir might have been located at a shallow level. Indeed, field evidence in both Cima Pape and Predazzo complexes attest to a magmatic contact between the intrusive rocks and the volcanic counterparts marked by a grain size transition (Sommavilla 1969; Casetta et al. 2018b). The onset of shallow crystallization conditions of the intrusive bodies is also confirmed by both clinopyroxene and amphibole–plagioclase barometry in intrusive rocks, which indicate a last crystallisation stage between 40–165 (Casetta et al. 2018a) and 100–250 MPa (Nardini et al. 2022). In good agreement, our lowest P estimates on augitic rims in the volcanic rocks resulted in 120–160 MPa ± 30–80 MPa, representing the system's last and shallowest crystallization level. Considering this, the pressure estimates in the augitic cores resulted in 150–250 MPa on average (5–8.5 km considering a ΔP/Δz of 29 MPa/km) and could confirm the probable shallow-crustal nature of the trachyandesitic magma, despite the large range on the P estimates.

Diopsidic cores of type B crystals were instead formed in the more primitive trachybasaltic melt. Textural features coupled with barometric estimates indicate a deeper formation than the diopsidic bands. Their irregular resorbed and spotted/patchy texture indicates recycling dynamics (Streck 2008; Jerram et al. 2018; Di Stefano et al. 2020) highlighted by the extensive (but incomplete) diffusion, suggesting an increased residence time of the spotted and patchy-zoned cores. In fact, some patchy-zoned cores have lower Mg# than the typical diopsidic domain but still have a high Cr content, due to its slower diffusion rate compared to Fe–Mg interdiffusion (e.g. Fig. 4b) (Cherniak and Dimanov 2010). Notably, the spotted/patchy-zoned and mottled cores record the highest P (300–500 MPa; ⁓10–17 km), similar to those resulting from the primitive best-preserved clinopyroxenitic nodules from the Latemar area. This analogy suggests that the clinopyroxenitic nodules could represent fragments of a diopsidic mush partially cannibalised by the mafic input that remobilised some of the crystals into the shallower parts of the plumbing systems (Fig. 7). These diopsidic antecrysts were then resorbed and partly dissolved, creating the variably developed mottled textures due to juxtaposition of dissolved and resorbed antecrysts (Eggins 1993; Bennett et al. 2019; Palummo et al. 2021). After being overgrown by augite, once brought to the shallower magmatic batch, some recycled Type B antecrysts were then mantled by one or multiple diopsidic bands, testifying for the subsequent mafic inputs (Type B1 and B2). Unlike the cores, these latter have euhedral shapes and sharp edges without any resorption, they often have a higher Mg# and record a lower P supporting the different formation environments for the two features. More frequently, diopsidic antecrysts were rimmed by augitic composition with oscillatory zoning as evidence of the convective dynamics in the trachyandesitic shallower reservoir (Type B0 crystals) (see Fig. 2b, c, Table 2).

In addition to the different magmatic environments, clinopyroxene texture discriminates the level of undercooling of the melt (ΔT = Tliquidus – Tcrystallisation): euhedral textures indicate very low rates of undercooling, while disequilibrium textures like dendritic or skeletal shapes witness higher undercooling rates (Mollo et al. 2010, 2011; Masotta et al. 2020; Ubide et al. 2021). The presence of sector-zoned pattern in Type C clinopyroxene combined with other zoning patterns (see Fig. 2a, d) also provides information on the kinetic conditions indicating a low rate of undercooling, in which elements like Si, Mg, Al and Ti are preferentially partitioned into the hourglass or prism sector of the crystal (Ubide et al. 2019a). The high number and evolved nature of the melt inclusions suggest that their entrapment occurred in the last stages of crystallization, when the undercooling was higher, and the ability to crystallise was easier. The Ti enrichment in clinopyroxene from lava samples from Predazzo and Mt. Monzoni could be explained by the differences in the undercooling rates: indeed, the Al–Ti enrichment is the result of crystallisation under higher degrees of undercooling (Mollo et al. 2010, 2011; Ubide et al. 2019a). This is confirmed by the absence of Type C crystals (resulting from low degree of undercooling) from lava samples (they are only present in dykes) and the higher number of inclusions in the clinopyroxene cores. Occasionally, sector-zoned crystals are mantled by diopsidic bands after a mafic injection as documented by the rare Type C1 crystals.

Timing the pre-eruptive magma dynamics

Complex zoning patterns in clinopyroxene populations from different volcanic centres indicate that crystallisation and storage took place in two different magmatic environments: a colder low-Mg#, low-Cr and a hotter high-Mg#, high-Cr environment. The latter is represented by the mafic input periodically refilling the evolved and shallower reservoir and is accountable for the formation of the diopsidic bands. The pre-eruptive timescales were calculated using the NIDIS (non-isothermal diffusion incremental step) of Petrone et al. (2016). This method considers how the different temperatures of each magmatic environment affect the diffusion rate, to track the residence time (Δt) from mafic injection to eruption in Type A1 and A2 crystals and the residence time in the shallower reservoir of Type B crystals. Diffusion coefficient is calculated with the Arrhenius equation: \(D={D}_{0}{e}^{\frac{-\Delta H}{RT}}\) , where R is the gas constant (R = 8.3145 J mol−1 K−1), T is the temperature (in K), and D0 and ΔH are the pre-exponential factor and the activation energy, respectively (Dimanov and Sautter 2000). The calculated total residence time (Δt) corresponds to the time elapsed from the first formation of the diopsidic domain (i.e. diopsidic band) to the thermobarometric shutdown of the system, i.e. the eruption. According to the NIDIS model, Δt is calculated using a backward approach from the last-formed compositional boundary (Δt2) inward (i.e. Δt1) (Petrone et al 2016, 2018). For this aim, the diffusion profiles are modelled by exploiting the difference in the grey values in BSE images between the diopsidic (low grey values) and the augitic domain (high grey values) (see diffusion profiles in Supplementary Figure S4). The greyscale level of clinopyroxene is calibrated against the Mg# determined by the electron microprobe.

In the case of Type B0 crystals, resolving the diffusion profile between diopsidic core and augitic mantle/rim provides the minimum time from the remobilisation of the diopsidic cores from the diopsidic magma mush and the eruption, i.e. the residence time of the crystal in the main shallow ponding zone (Δt2; Fig. 8b)). Assuming that the remobilisation of the diopsidic antecrysts is due to mafic input that permeates the diopsidic mush, this time could be interpreted as the last mafic input-to-eruption timescale. The other crystal zoning patterns require more complex modelling. In fact, when a diopsidic band is present (i.e. Type A1), two compositional boundaries (i.e. core–band and band–rim) occur. Following the approach of Petrone et al. (2016, 2018), at the band–rim boundary, we calculated the Δt2 which represents the diffusion time spent in the low T, low-Mg# melt domain, i.e. the rim diffusion timescale. For the core–band boundary, we calculate two different timescales: Δt3, which is a pseudo-timescale assuming the core–band profile has diffused at low T to reproduce the same band–rim diffusion profile (i.e. same curvature of the curve) as Δt2; Δt4 which is instead the pseudo-timescale assuming it diffused entirely at high T. Since the core–band boundary has diffused both in high and low T, the real timescale of the core–band boundary (Δt1) is then obtained with the difference Δt4 − Δt3 (Petrone et al. 2016). The diffusive time elapsed between the arrival of the mafic injection and eruption is then given by the sum of Δt1 (i.e. diffusive time in the high-Mg# magmatic environment) and Δt2 (i.e. diffusive time in the low-Mg# magmatic environment) (Fig. 8b). Since it indicates the residence time in the high-T magma, Δt1 provides timescales needed to homogenise the diopsidic melt into the more voluminous augitic one (i.e. mixing timescale; Petrone et al. 2018).

Fig. 8
figure 8

a Summary of the timescales and related errors (see text) calculated from the injection of the mafic magma to the eruption in the four centres. b Grey value diffusion profiles of two representative clinopyroxene crystals from the Dolomites calculated inside the red area of the associated BSE-SEM images (see Supplementary Figure S4 for all modelled profiles). The first example comes from Cima Pape and describes the computations exploited to reconstruct the time from injection (formation of the band) to eruption. Below, diffusion chronometry on the crystal from Mt. Monzoni expresses the total residence time of the diopsidic antecryst in the evolved magma (at the evolved magma conditions of T) which corresponds to the time from the mafic input to the eruption. The blue line refers to Δt4, and the red line refers to Δt2 and Δt3 (see text for details)

Full size image

For the calculations, we have preferentially taken into consideration transects normal to the {100} sector, perpendicular to the c-axis (e.g. {010}), considering that in pyroxene crystals the c-axis is kinematically controlled by more rapid crystal growth and faster diffusion velocity compared to the other two crystallographic directions. Indeed, Fe–Mg diffusion along the c-axis in pyroxene is faster than in the other directions (Allan et al. 2013; Müller et al. 2013; Dohmen et al. 2017).

Considering that the slow diffusing elements (Ca and Al, see Supplementary Fig S3) show sharp change that mimics the Mg# profile, we can safely assume that the observed Fe–Mg compositional profile is a diffusion profile; therefore, the inner boundaries should always be smoother than outer boundaries, because the latter have a shorter time to diffuse than the core–band boundary. This situation is occasionally not fulfilled for the clinopyroxene considered in this study. Few clinopyroxene crystals clearly describe a steep diffusion profile at the core–band boundary and a smoother profile at the band–rim boundary. As already demonstrated by Petrone et al (2018) at Stromboli volcano, this apparent anomaly could suggest that the band–rim boundary records the incomplete homogenisation between the two melts (homogenisation time longer than residence time), thus capturing the ongoing magma mixing (“anomalous profiles”, Petrone et al. 2018). This process is emphasised by the frequent occurrence of double plateaux in some diopsidic cores and bands, where a hybrid composition is present (Figs. 3d and 4c). In these cases, mixing between the two melts was slower than the growth rate, which was nearly instantaneous, thus enabling the preservation of a double plateau in the external diopsidic–augitic compositional boundary. To avoid biases in the timescale calculations (see Couperthwaite et al. 2021), we excluded these types from the diffusion chronometry modelling.

The results of the computations and related errors are shown in Fig. 8; the diffusion profiles and the timescale for every crystal investigated are present in Supplementary Figure S4. The error propagation at 66% confidence level on the time estimate is reported considering the uncertainty on \(\surd 4Dt\) parameter (2σ) (fitting parameter; see Petrone et al. 2016) and on temperature (1σ). The calculated timescales indicate that the shortest timescales (< 1 year) are extracted from type B0 clinopyroxenes in Mt. Monzoni. Here, the average residence time is 1.5 years, while the range of timescales is from 0.1 to 5 years with 60% of the computed crystals returning residence times of less than a year. These timescales are much shorter than those from Predazzo and Cima Pape crystals, which record timescales from years to decades, or even up to 80 years in the largest crystals. In the Sciliar clinopyroxene population, the only two investigated crystals record pre-eruptive timescales of 11 and 34 years similarly to Predazzo and Cima Pape (Fig. 8a). Overall, the calculated timescales reflect different timing for the same magmatic process between Mt. Monzoni and the other three centres. Considering that all the zoned crystals in Mt. Monzoni products are Type B0, the short timescales recorded in these samples witness the residence time spent by diopsidic antecrysts in an evolved magmatic environment after mafic injection. This process brought them to a shallower level.

Effect of the mafic input in the plumbing systems beneath the Dolomites area volcanoes

After injection into the evolved magma reservoirs, mafic melts experienced mixing and homogenisation. The mixing dynamic was frequently short (Δt1 < 1 year) and occasionally recorded in the crystals where a double plateau is preserved at the diopside–augite interface (see Figs. 3d and 4c). The presence of this hybrid composition in the crystals indicates that the growth rate of these clinopyroxenes was fast enough to register the ongoing mixing between the two melts, despite its short timescale, as already documented at Stromboli volcano (Petrone et al. 2018).

The presence of diopsidic antecrysts is predominant in Mt. Monzoni samples where, conversely, diopsidic high-Mg–Cr bands were not found. This suggests that the Mt. Monzoni centre was characterised by slightly different plumbing system dynamics. The sole presence of Type B0 and C crystals among the zoned population and the complete absence of Type A1, A2, B1 and B2 crystals suggest that the mafic input was capable of recycling diopsidic antecrysts more effectively. This implies an efficient mechanism of mush disruption and cannibalisation, in which the antecrysts are continuously remobilized and transported by the mafic magmas permeating the diopsidic mush. These diopsidic antecrysts were then brought into the shallower trachyandesitic reservoir and coated with thin (⁓10 µm) oscillatory augitic rims (Δt < 5 yrs; Fig. 7b). Such thin rims indicate that antecrysts resided for short time in the trachyandesitic ponding zone, suggesting a more efficient triggering nature of the mafic input in this complex compared to the other coeval centres. These rims prove that the last event in the plumbing system before the eruption was the diopsidic antecryst remobilisation. Type B0 crystals from Predazzo, Cima Pape and Sciliar present thicker augitic rims associated with longer timescales (i.e. Δt > 10 yrs), sign of an efficient long-lasting storage in the more evolved melt. The oscillatory zoning in the augitic domain testifies to crystallisation in a dynamic environment, characterised by convection possibly enhanced by mafic replenishment (Di Fiore et al. 2021; Petrone et al. 2022).

Conversely, in the other centres, zoned Type A crystals (i.e. A1 and A2) are more common than Type B crystals, highlighting a difference with respect to Mt. Monzoni. Furthermore, the significant presence of Type A2 clinopyroxene, which testifies to multiple mafic inputs, suggests that mafic recharges did not trigger eruptions, contrary to what was observed in Mt. Monzoni clinopyroxene. This is matched with the diffusion profiles modelled in the Predazzo and Cima Pape crystals that resulted in longer timescales. The non-triggering nature of the mafic replenishment could suggest that eruption was possibly driven by the tectonic activation of the systems related to the distensive–transtensive dynamics present in the Dolomites area (Doglioni 1987). A clue of this hypothesis is found in the texture of the volcanic rocks which show complete absence of glass, crystalline matrix, high PI, large crystals, and frequent and large glomerophyres. Considering the short time span of the mafic volcanism in the Dolomites area (< 0.9 Ma; Storck et al. 2019), the differences in timescales and texture between Predazzo and Mt. Monzoni clinopyroxene populations suggest a separate nature and dynamics of the two systems, contrary to previous models (Abbas et al. 2018). Another evidence of this is the higher abundance of biotite in intrusive rocks from Mt. Monzoni with respect to those from the Predazzo pluton (see Bonadiman et al. 1994; Casetta et al. 2018a, b). Analogously, the highest number of zoned crystals and the highest values of Mg# (up to 91) and Cr2O3 content (up to 1.2 wt%) in the diopsidic domain found in Cima Pape suggest a larger amount of mafic input in the evolved reservoir, probably due to local pathways allowing the primitive magma to reach the shallower reservoir. Altogether, these features demonstrate that, besides similar architecture and dynamics, each single volcanic centres show unique intensity and frequency of plumbing system’s processes.

A model for active volcanoes: a comparison with Stromboli

The well-preserved Middle Triassic complexes of the Dolomites area are easily comparable with currently active magmatic systems, where the mafic-recharge dynamics are often recorded by pyroxene crystals (Morgan et al. 2004; Saunders et al. 2012; Ubide and Kamber 2018; Mangler et al. 2020). The characteristics of the studied clinopyroxenes are akin to the clinopyroxenes hosted by the volcanic products of Stromboli (Petrone et al. 2018, 2022; Di Stefano et al. 2020). At Stromboli, the plumbing system is typified by periodic replenishment of a shallow degassed highly porphyritic magma batch (Hp) by mafic volatile-rich low porphyritic magma (Lp) (Métrich et al. 2001; Francalanci et al. 2012; Petrone et al. 2006, 2018; Pichavant et al. 2011; Ubide et al. 2019b; Di Stefano et al. 2020). Mafic recharge induces the crystallisation of diopsidic bands (Mg# 77–91) and rims around augitic cores (Mg# 70–76) present in a shallower ponding zone and the recycling of diopsidic resorbed antecrysts (Francalanci et al. 1999; Bragagni et al. 2014; Di Stefano et al. 2020). With time, the Stromboli plumbing system has changed its internal structure and, as a result, the two compositional clusters have modified their organisation within the zoned crystals: the frequency of the zoning pattern has changed from diopsidic band-dominated crystals (Post-Pizzo activity, ∼1.7–1.5 ka; Petrone et al. 2018) to antecryst prevalence (2003–2017 activity; Di Stefano et al. 2020). Then, in the paroxysms of June–August 2019, the antecrysts are extremely rare and the common zoning type is represented by clinopyroxene with diopsidic rims, testifying that the mafic inputs represent a direct trigger of eruptions in a rejuvenated plumbing system after efficient cannibalisation of the existing mush (Petrone et al. 2022). Accordingly, we speculate that in the Mt. Monzoni system the mafic recharge had a stronger remobilising effect on previously crystallized diopsidic cores carried and recycled by mafic recharge, leading to a situation similar to the 2003–2017 activity of Stromboli (Di Stefano et al. 2020), while in the other systems (Predazzo, Cima Pape, Sciliar), the mush disruption capability of the mafic input was limited, and generated crystal textures similar to those found in the Stromboli Post-Pizzo period (Petrone et al. 2018).

Summary and conclusions

In this work, a detailed textural/compositional study on clinopyroxene crystals in Mid-Triassic lavas and dykes from the Dolomites area (Predazzo, Mt. Monzoni, Cima Pape and Sciliar) was performed to reconstruct the dynamics and architecture of the plumbing systems beneath the ancient magmatic complexes. Detailed modelling of the intensive parameters of the magmas feeding each volcanic centre and application of diffusion chronometry models led us to the following findings:

  • Textural, compositional and thermobarometric features indicate that mafic replenishment processes were common in the volcanic centres of the whole Dolomites area, namely, Predazzo, Mt. Monzoni and Sciliar volcano-plutonic complexes. Estimates of magma storage conditions record similar architectures in the four centres, where a shallow-crustal (5–8.5 km) trachyandesitic reservoirs (Mg# = 45; T = 1044–1118 °C) was periodically replenished by hotter and more primitive trachybasaltic melts (Mg# = 56; T = 1056–1170 °C).

  • The comparable composition and estimated magma storage conditions between diopsidic antecrysts and the clinopyroxene in the clinopyroxenitic nodules from the Latemar area suggest that the latter could represent the diopsidic mush source of the primitive cores found in lavas and dykes. Estimated storage pressures indicate that the diopsidic mush ponded in the mid crust (10–17 km) periodically permeated by the mafic input which brought the diopsidic antecrysts from the deeper mush into the evolved shallower storage.

  • Through NIDIS modelling of the diffusion profiles in the clinopyroxene crystals (Petrone et al. 2016), we calculated injection to eruption timescales of the crystals ranging from a few months (Mt. Monzoni) to years and decades (Predazzo, Cima Pape and Sciliar). Through this modelling, we discriminated different degrees of eruption trigger efficiency of the mafic input among the volcanic centres.

  • Bearing in mind that the volcanism in the Dolomites area lasted < 0.9 Ma, the different textural patterns and timescales recorded by Fe–Mg diffusion in clinopyroxene phenocrysts suggest that Predazzo, Cima Pape and Sciliar plumbing systems were not triggered by the mafic input, while at Mt. Monzoni the mafic input was able to largely destabilise the systems toward the eruption. At the same time, the evolved reservoir was more efficiently stirred, resulting in more effective crystal cannibalisation. These features suggest that Mt. Monzoni was probably not linked to the Predazzo complex also accounting for the slight differences in the paragenesis of intrusive rocks (Bonadiman et al. 1994; Casetta et al. 2018a, 2021). This work introduces a solid starting point for linking the magmatic processes registered in the volcanic record and the intrusive rocks which give information on the system storage conditions, internal dynamics and long-term history.

  • The broad similarities in textures and composition between the clinopyroxene from the Mid-Triassic centres and those from the active volcanic systems (e.g. Stromboli; Petrone et al. 2018, 2022; Di Stefano et al. 2020) give a great opportunity to test and improve the models used in the currently active volcanoes, thanks to the remnants of the long-term history of the plumbing system preserved in the Dolomites area. This highlights the potential of studying ancient, entirely exposed volcanic systems for interpreting the feeding system processes working beneath active volcanoes.