Introduction

About 75–83% of the bulk volume of magma ascending from the asthenosphere stalls out at lithospheric discontinuities (Crisp 1984). Magma ascent through these discontinuities versus indefinite stagnation depends on the difference in physical properties between the rising magma and the surrounding rock, the active tectonic stress, lithostatic pressure, and the difference in temperature and composition between magma and country rocks (Ryan 1994; Watanabe et al. 1999; Menand 2011). The development of a magmatic system is therefore influenced by the variations in the magma rheological conditions along the whole extent of the system. While the shallow magma system includes small crystallised bodies and is free of magma, the deep magma system might accumulate magma volumes. These volumes may evolve through partial to total crystallisation to generate gabbros, crystal mushes and ultramafic cumulates (Cashman et al. 2017; Sparks et al. 2019). Hundreds of thousands to millions of years of these processes result in the formation of several kilometre-thick ultramafic layers at the base of the crust, akin to the Canary Archipelago (Hansteen et al. 1991, 1998; Klügel et al. 2022), La Reunion Island (Boudoire et al. 2019) and Mayotte Seamount (Mittal et al. 2022).

The prompt detection of a deep magmatic intrusion, prone to originate a surface eruption, relies upon the correct XYZ spatial location of the associated seismicity. At many volcanic islands, in the absence of detailed models of Vp and Vs distribution with depth, this parameterisation relies upon general models of oceanic crust structure. In these models, the crust-mantle boundary is defined from a major jump in P-wave velocities, i.e., from 6.6–7.3 km/s to 7.6–8.3 km/s (Rabbel et al. 2013), which should mark the passage from mafic intrusions (ρ ~ 2800 kg/m3) to ultramafic lithologies (ρ > 3300 kg/m3). In the central-north Atlantic archipelagos (Azores, Madeira, Canary, Cape Verde), this jump is found in a depth range between 12 and 15 km (Banda et al. 1981; Lodge and Helffrich 2006; Matias et al. 2007; Pim et al. 2008; Ranero et al. 1995). However, the constitution and thickness of the crust below oceanic islands with long-lived magmatism (~ 39 Ma for the Azores) are complicated by the effect of repeated magma intrusions and crystallisation events, which generate several jumps in Vp/Vs (e.g., Zanon et al. 2020; Silveira et al. 2010). Each of these jumps marks a shift at depth to rocks with different rheological properties and ρ > 3100 kg/m3 (clinopyroxene and olivine mixtures), which, therefore, are misinterpreted as the mantle.

In ultra-slow spreading systems such as the Azores Islands, the volumes of ascending melt through the crust are reduced, being preferentially accumulated at its base, and cooling is so efficient that a steady-state melt lens cannot be maintained (Lissenberg et al. 2019). Melt crystallisation promotes the formation of a range of crystalline lithologies, ranging from gabbro to syenite and ultramafic cumulates. This means that the progressive accumulation of un-erupted magmas contributes to generating “new crust layers”, which are geophysically indistinguishable from the mantle due to its mineralogical constitution and interbedded with portions of the lithospheric mantle. At the same time, magma intrusions in the upper lithospheric should act as metasomatising agents to change mantle composition and rheological characteristics. The real boundary between the crust and the lithospheric mantle is, therefore, difficult to determine and is dispersed in a transition zone with a variable thickness (e.g., Benn et al. 1988; Boudier and Nicolas 1995; James et al. 2003; Ghosh et al. 2014).

Therefore, the application of general models is responsible for errors in the spatial location of seismicity. The magnitude of these errors is proportional to the thickness of the crust and its complexity. A correct definition of the crust-mantle transition is, therefore, fundamental for seismovolcanic monitoring activities.

The Azores archipelago in the central-north Atlantic Ocean covers an area of ~ 52,000 km2 and straddles the Mid-Atlantic Ridge. These islands lie on top of a thick crust (Searle 1980; Detrick et al. 1995; Luis et al. 1998; Cannat et al. 1999; Gente et al. 2003; Silveira et al. 2010), generated by the excess of magmatism occurring from in the Miocene (Luis and Miranda 2008). The frequent seismicity that characterises most of the islands is associated with the geodynamics of the area and has an almost exclusive tectonic character. The availability of eruptible mafic magmas, which may rapidly ascend from the deep lithosphere, makes these oceanic islands suitable sites to investigate the depth of crust-mantle transition and comprehend the effect of tectonic stresses on the architecture of magma systems. To fulfil this aim, we used:

  • New microthermometric data on CO2-rich fluid inclusions in the mineral assemblage from high porphyritic mafic lavas and tephra in complement to literature data;

  • Seismicity recorded by the seismic monitoring network of the Azores during the last 22 years of monitoring activity.

Geological setting

The Azores region, in the central-north Atlantic Ocean, between 25 and 32° longitude W and 36 and 40° latitude N, includes nine islands and seamounts of volcanic origin (Fig. 1). They are distributed along 600 km across the boundary between three plates and are straddling the Mid-Atlantic Ridge (MAR), which runs along a N20° direction. The islands of Flores and Corvo are on the North American plate, while all the others are between the ultra-slow Terceira Rift to the north and the “East Azores Fracture Zone” transform fault to the south. The border between Eurasia and Nubia is diffuse. The Azores region east of the MAR adjusts the stresses generated by the rotation of the Nubia plate relative to Eurasia through a series of subparallel spreading ridges oriented along a general N120° direction (e.g., Vogt and Jung 2004; Borges et al. 2007; Marques et al. 2013; Hildenbrand et al. 2014; Miranda et al. 2014; Fernandes et al. 2018). The northernmost of these spreading ridges is the Terceira Rift, which hosts the islands of Graciosa, Terceira and S. Miguel. Its spreading velocity away from Eurasia is 2 ± 1 mm/yr (Marques et al. 2013). To the south and close to the MAR, there are the spreading ridges of S. Jorge Island and that one hosting the islands of Faial and Pico. For this latter ridge, the spreading velocity is 3.5 ± 0.5 mm/yr, higher than at the Terceira Rift.

Fig. 1
figure 1

Location of the archipelago and the geodynamic context of this area of the Atlantic Ocean. The thick red lines are the Terceira Rift which marks the northern boundary of the Azores region and the transform fault “East–West Azores Fracture Zone”. The thinner red lines mark the São Jorge and Faial-Pico spreading systems. The thick black line is the Mid-Atlantic Ridge. Panels from b to f show sampling sites on the islands of S. Miguel, Terceira, Pico-Faial, Graciosa and S. Jorge

Full size image

The succession of fissure zones and central volcanoes in the islands evidences a strong relationship between volcanism and tectonics. Geophysical data analysis and field studies have recognised strike-slip faults at Faial (N75° system; Marques et al. 2014), Pico (N60° system; Costa et al. 2014) and S. Miguel (N45° to N80° systems; Sibrant et al. 2016). The intersection between these faults with the regional N120° fissure systems (Fig. 1) enhanced a rapid magma withdrawal from the deepest reservoirs that occurred without long-term ponding and differentiation within the crust (Zanon et al. 2020) and promoted discrete CO2 degassing (e.g., Viveiros et al. 2010; 2017).

In the area of the Azores, three cycles of diffuse submarine fissure volcanism began about 39 million years ago and lasted up to the Holocene (Beier et al. 2015). This activity led to the formation of a plateau with regional extension. The island of Santa Maria, south of S. Miguel, formed starting from ~ 6 Ma (Marques et al. 2020). S. Jorge and Flores islands ~ 1.9 and ~ 2 Ma (Marques et al. 2018; Hildenbrand et al. 2018), S. Miguel, Faial and Graciosa emerged between 850 and 700 ka (Sibrant et al. 2014; 2015a; Hildenbrand et al. 2012). As regards Terceira, Corvo and Pico, there are no dates of the basal lavas. In the latter case, a minimum age of 186 ± 5 ka has been provided by Costa et al. (2015). Holocene magmatism characterised most of the volcanic systems on all the islands and seamounts east of the MAR, except Santa Maria. However, the only relevant geophysical activity in the region is represented by recurrent seismicity, mostly of tectonic origin, even though small magmatic intrusions might be associated.

Samples collection

In this work, CO2-rich fluid inclusions from 44 samples of volcanic products younger than 20 ka, collected along fissure zones and at central volcanoes from Graciosa, Faial, Pico, Terceira, S. Jorge and S. Miguel have been studied (Table 1). Samples from much older volcanic areas in the eastern part of the islands of S. Jorge and S. Miguel and from Flores and Santa Maria have also been included for comparison. Figure 2 shows sample collection sites for each island, also including those reported in the literature (Zanon 2015; Zanon and Pimentel 2015; Zanon and Frezzotti 2013; Zanon et al. 2020; Métrich et al. 2014; Zanon and Viveiros 2019), which have been considered for completeness due to dense fluid inclusions.

Table 1 Sample data location, type and age
Full size table
Fig. 2
figure 2figure 2

Digital elevation model of the islands considered in this study, showing sampling sites. In these maps, orange dashed lines represent fissure zones and blue lines are the main faults. Names mark the location of central volcanoes referenced in this paper. White stars are sampling sites from earlier works whose microthemometric data are available. Red stars represent new sampling sites

Full size image

Analytical procedures

Seismicity

Data for the present study were recorded by the seismic permanent monitoring network Centro de Informação e Vigilância Sismovulcânica dos Açores (CIVISA). This permanent network includes 33 short-period and 4 broadband stations deployed on seven out of nine islands. The initial database contained 73,928 earthquakes recorded from 2000 to June 2022 and located based on the standard seismic 1D velocity model (Senos et al. 1980). Figure 3 shows the spatial distribution of the events in the region. These events interested the whole region, from the MAR to S. Miguel. No seismicity was recorded below Flores and Corvo islands. The dataset was reduced to 23,418 earthquakes resulting from the selection of events recorded from at least 8 stations, with a Root Mean Square (RMS) ≤ 0.3 s and with latitude, longitude and depth errors ≤ 5 km. For this study, seismicity was separated into two clusters: events sourced below Faial, Pico, S. Jorge, Graciosa, Terceira islands and surrounding submarine zones, hereby named as Central Group, and events originated below S. Miguel and Santa Maria islands and the Hirondelle Basin and Dom João de Castro Seamount, hereby named the Oriental Group. The split of the seismicity aimed to assess the influence of the Moho depth on the velocity model used for each group and played a main role in the final location of the seismicity. Therefore, for the Central Group, the boundary between the crust and the lithospheric mantle was considered at 20 km depth, and for the Oriental Group at 30 km depth.

Fig. 3
figure 3

Map of the distribution of seismic events recorded during 22 years in the Azores region. Blue lines mark active main transtensive faults

Full size image

The data from the Central Group comprised 7579 earthquakes that were relocated using the HYP software, a modified version of HYPOCENTER (Lienert and Havskov 1995), integrated with the SEISAN 11.0 analysis software (Havskov and Ottemöller 1999), and the layered velocity model obtained by Senos et al. (1980) considering the Moho at a depth of 20 km (see further). From this relocated dataset, the earthquakes with an error location ≤ 5 km were selected, reducing the number of considered events to 7137. This double selection guaranteed that only events with a high-quality location were used.

The same procedure was applied to the data from the Oriental Group, considering the depth of the boundary between the crust and the lithospheric mantle at 30 km (see further). The original number of events of 15,839 was reduced to 15,427 after the selection of events with location errors ≤ 5 km. For the seismicity recorded along the MAR, the selection criteria needed to be reduced (RMS ≤ 0.6 s), otherwise, it would not be visible. Because of the seismic network configuration and the epicentral distance to the MAR, the residual errors in the location of these earthquakes are larger.

CO2-rich fluid inclusions

Lava samples were first observed under a binocular to identify the crystal size suitable for the analyses and then crushed with a jaw crusher aperture corresponding to the largest crystal size. About 150/200 olivines and 100/150 clinopyroxenes have been hand-picked under the stereomicroscope from both pyroclastic samples and crushed lavas. They have been embedded in epoxy and doubly polished up to a final thickness of ~ 120 to 80 μm and checked for fluid inclusions. The micro-thermometry study has been carried out on a Linkam MDSG600 heating–cooling stage at the Instituto de Investigação em Vulcanologia e Avaliação de Riscos in Ponta Delgada (IVAR), Portugal. The stage has been calibrated according to synthetic fluid inclusion standards of pure CO2 and H2O. Melting and homogenisation temperatures are reproducible to ± 0.1 ℃, with heating rates between 0.2–0.5 ℃/min. The densities of the CO2 inclusions have been calculated based on the Eqs. (3.14) and (3.15) of Span and Wagner (1996) and corrected for water content (H2O:CO2 = 1:9) following Hansteen and Klügel (2008). Isochores for CO2–H2O inclusions have been determined based on the equation presented in Sterner and Bodnar (1991). The error of measurement of fluid properties is ± 0.1 ℃, which determined an estimated maximum density error of ± 1.0 kg·m−3 in olivine-trapped fluid inclusions after density correction. The resulting error in pressure calculation is ± 1.4 MPa at 1170 ℃, which determines an error of ± 20 m in depth definition for the densest inclusions at Flores and Santa Maria, ± 40 m at Pico, Faial, Graciosa, Terceira and S. Jorge, ± 50 m at S. Miguel.

Data description

Seismicity

Seismic data highlight spreading activity at the rift systems in the islands and active seismogenic zones, such as at the W and NE of Faial Island or the Água de Pau volcano at S. Miguel Island. The island of Flores is inactive. Most of the recorded earthquakes are volcano-tectonic events, with clear P- and S-phases onsets. Rare hybrid and long-period events were detected associated with the Água de Pau volcano. Seismic magnitudes in the archipelago are moderate to low, with rare energetic events up to ML5; such was the case of the 1998 CE Faial earthquake or the 2007 CE earthquake between S. Miguel and Santa Maria islands at Formigas Islets. The fault plane solutions obtained for these two strongest earthquakes show strike-slip to normal fault mechanisms, characteristics of trans-extensional regimes, which is expected considering the Terceira Rift and MAR domains (Silva et al. 2020).

A selection of recorded filtered seismicity is presented in the Supplementary Table 1. Beneath the MAR, recorded seismicity is not as shallow as below the islands closer to the MAR, where most of the relocated hypocentres are within the first 15 to 20 km depth of the crust with single shocks. Regarding the Central Group, swarms of earthquakes reaching 18–20 km depth are associated with the seismogenic zone east and west of Faial. The present spreading occurring in the central area of S. Jorge (2022 CE crisis) is generating seismicity focused at 10–12 km, with scattered events up to 36 km deep. Other sparse, deep events (up to 30.4) occurred below Graciosa and between Pico and Faial. Clustered shallow depth microseismicity characterises areas west of Faial (around 5 km to sub-superficial), as well as the central area of Terceira (between 5 and 10 km deep).

At the Eastern Group, beneath S. Miguel Island and surrounding areas, hypocentres are within the first 15 km depth of the crust, except for the Dom João de Castro Seamount and the Hirondelle basin (Fig. 1), where they range between 25 and 30 km depth. At S. Miguel, most micro seismicity is associated with the Água de Pau volcano, in the central part of the island, between 0 and 15 km depth, with scattered deeper events (Fig. 3).

Sample characteristics

Lavas are porphyritic with a seriate intergranular to intersertal texture and crystal contents within the range of 20–70 vol% (phenocrysts and micro-phenocrysts). Poorly porphyritic-to-aphyric lavas (< 10 vol%) have not been considered in this study. The mineral assemblage is given by olivine and clinopyroxene, locally joined by plagioclase (Fig. 4a). Minor phases are apatite, Fe-Ti oxides (Ti-magnetite and/or ilmenite), and occasional amphibole (kaersutite). The same minerals, but with a dominant amount of plagioclase over mafic phases and Fe-Ti oxides, can be found in the groundmass. Highly porphyritic rocks are common at central volcanoes and less frequent at fissure zones.

Fig. 4
figure 4

Photomicrographs of typical studied lava samples and fluid and melt inclusions. High porphyritic lava from a cinder cone on the southeast coast of Terceira Island is shown in (a). Large plagioclase phenocrysts are evident in the mineral assemblage, besides olivines and clinopyroxenes. An ultramafic cumulate is present. High porphyritic lava from the Pico volcano is represented in (b) and shows the occurrence of very large crystals (megacrysts) of clinopyroxenes and large crystals of olivines. c Type-I fluid inclusions (diffuse) trapped in olivine in a lava balloon from the 1998–2000 CE submarine eruption at Terceira. All inclusions show re-equilibration features. Small silicate melt inclusions are also visible. d Type-II fluid inclusions (trails) trapped in an olivine antecryst from the Sete Cidades volcano. All inclusions show re-equilibration features

Full size image

Fallout tephra contains juvenile porphyritic to hypocrystalline and holohyaline clasts and loose crystals. The mineral assemblage is the same as that found in the corresponding lavas.

Harzburgite fragments of variable size have been rarely found in three samples from Pico, Faial and S. Miguel. Other ultramafic lithologies (dunites, clinopyroxenites and wehrlites) with an adcumulate structure have been found at Pico and S. Miguel, together with plagioclase-free gabbros and syenites.

Studied olivine and pyroxene are phenocrysts (Øolivine ≤ 2 mm; Øpyroxene ≤ 5 mm) and megacrysts (Øolivine ≤ 4 mm; Øpyroxene ≤ 25 mm for pyroxenes–Fig. 4a, b). Megacrysts are common in lavas from the Pico and Sete Cidades volcanoes, where they are dispersed or accumulated in discontinuous layers, originating from the disruption of consolidated crystal mushes during the ascent of magma along tectonic discontinuities (Zanon et al. 2020).

Olivines are euhedral (polyhedral), subhedral or skeletal; clinopyroxenes are euhedral or subhedral and larger phenocrysts are characterised by oscillatory or rarer hourglass zoning. Pyroxene megacrysts always show disequilibrium growth features such as patchy zoning, resorbed cores, reaction rims and embayments (Fig. 4b). Plagioclase crystals (350–2200 μm in length) can be zoned or show boxy-cellular and sieve textures. Fe-Ti oxides are common as microphenocrysts (150–300 μm) in lavas. Rare large (up to 600 μm) euhedral Ti-magnetites have been occasionally found in tephra deposits.

CO2-rich fluid inclusions

Fluid inclusions have been found in mafic assemblages of both tephra and lavas and are differently distributed among the mineral phases: they occur in ~ 15 to 25% of olivines and ~ 5% of clinopyroxene phenocrysts and are absent in microphenocrysts. There are two kinds of inclusions based on their texture. Type I inclusions are infrequent, isolated or forming small clusters. Their morphology ranges from negative-crystal to rounded, with a maximum diameter of 35 μm. Re-equilibration features are common (Fig. 4c). Type II inclusions are abundant in both olivines and pyroxenes, where they form trails of variable length and thickness, which line healed fractures, terminate at or crosscut grain boundaries (Fig. 4d). No daughter minerals have been found in any inclusions.

At room temperature, fluid inclusions are either single-phase (L) or contain a vapour bubble (L + V). In a few cases, there are two components, with two liquid-rich nested bubbles (L1 + V1 + L2). Most inclusions froze between − 68 and − 89 ℃, and most of them showed instantaneous melting at − 56.6 ± 0.3 ℃, thus showing an almost pure CO2 composition of the trapped fluid. Lower melting temperatures (up to − 57.4 °C) were recorded in < 1% of the studied population and are justified by < 5% mol proportions of other gas species, which lower the melting temperature (Van den Kerkhof and Thiery 2001). However, as such a low amount of other species does not substantially change the interpretation of trapping conditions (Van den Kerkhof 1990), we did not take them into account in the calculation of the density. Final CO2 homogenisation occurs either in the liquid (ThL) or the vapour phase (ThV) up to < 31.1 ℃.

In large two-component inclusions, clathrate melting was detected starting at -11 °C. However, the full process was not accurately determined because of the poor transparency of the host crystals, and the original water content was not measured. It was therefore assumed to have a plausible maximum water content of 10 mol%, in agreement with Hansteen and Klügel (2008) and Dixon and Stolper (1995). Histograms of ThL and density values, recalculated for 10 H2O mole% of all measured fluid inclusions in olivines and clinopyroxenes, are shown in Supplementary Figs. 1 and 2. These histograms also allow a rapid comparison with literature data.

Discussion

Information from CO2-rich fluid inclusions barometry

Rheological and stratigraphic heterogeneities along the magma ascent path are responsible for a non-isochoric ascent path (Frezzotti et al. 2004) and result in the formation of a vertically extended magma system where magma accumulates only at depth. At these deep ponding stages, CO2 might degas, as this is highly incompatible as a dissolved phase. In addition, a magma system can be fluxed by CO2-rich fluids released from a deep source (Caricchi et al. 2018). Therefore, at depth, fluid CO2 can easily be trapped in mineral phases which are crystallising or re-equilibrating, thus recording the barometric condition of each event of ponding.

During ascent, as a function of the velocity and rate of the strain and the hardness of the host mineral, already trapped fluid inclusions may stretch, leak or decrepitate (Bodnar 2003). These processes reduce the original density of the trapped fluid. During prolonged magma ponding, the re-equilibration process affects all the fluid inclusions, which reset to the new pressure conditions of the magma ponding, thus losing memory of deeper entrapment conditions. As these processes may occur at any stage of the ascent (stationary ponding or dynamic ascent), the barometric information of a single set of re-equilibrated inclusions is difficult to interpret.

However, by comparing the fluid inclusion distributions from multiple samples representative of different magmas erupted over a short period from various volcanic systems, it is possible to identify density peaks that represent important stratigraphic discontinuities at depth, representative of magma ponding stages (Supplementary Figs. 1 and 2).

Fluid inclusions in clinopyroxenes record lower densities than those in olivines (Table 2), as their structure facilitates inclusion stretching (Hansteen and Klügel 2008; Hildner et al. 2011). The only exception is represented by literature data for the recent eruptions in the central area of the Rosais-Manadas fissure zone at S. Jorge. The highest densities (ρ), recalculated for a possible 10% water content, are 1086 kg∙m−3 at Santa Maria, 1038 kg∙m−3 at Flores, 940–1003 kg∙m−3 at S. Miguel, 764 and 916–940 kg∙m−3 at S. Jorge. Below the other islands of the Central Group, the highest density values are around 829 ± 9 kg∙m−3 (Supplementary Table 2).

Table 2 Microthermometric database of analysed fluid inclusions
Full size table

Pressures of fluid trapping were derived from isochore distribution in the P–T space at the trapping temperature (Table 2) from literature data (Zanon and Frezzotti 2013; Zanon and Pimentel 2015; Zanon 2015). Regarding the studied volcanic systems at Flores, Santa Maria, S. Jorge and S. Miguel, it was assumed that they had the same trapping temperature as already studied volcanoes nearby, also because of the similarities in petrographic characteristics and proximity. The effect of the incertitude on the temperature calculation has a negligible effect on the calculation of the pressure (e.g., Hansteen and Klügel 2008). As an example, a difference in temperature of ± 10 °C does not significantly affect the resulting pressure values, which, for the interval of densities under consideration, vary only by ± 5 MPa in the worst case.

The resulting pressures are overall variable with the distance from the MAR and with the age of the volcanic system and range from 900 and 841 MPa below the oldest volcanic complexes at Santa Maria and Flores to 501 MPa below the recent Pico volcano. Large variability is shown at S. Miguel (from 550 to 807 MPa), with the highest values, very similar to each other, being recorded below the two fissure zones. Barometric information is quite similar below the islands of the Central Group, except for S. Jorge Island, which is the oldest of the group.

Deep magma accumulation zones below the Azores volcanoes

The lithological characteristics of the deepest magma system of a volcano change with time.

Past studies conducted at the Canary Islands, Madeira Island and Fogo volcano (Cape Verde) suggested that the deepest roots of their magma systems are still in the mantle, as some Holocene eruptions transported mantle and ultramafic cumulate xenoliths to the surface, collected from similar depths (e.g., Hansteen et al. 1991, 1998; Klügel et al. 2009, 2022; Hildner et al. 2011, 2012; Schwarz et al. 2004). A mature deep structure of a magma system probably consists of vertically stacked magmatic intrusions with different degrees of crystallisation, ultramafic crystal mushes interbedded by metasomatised peridotite layers. This brings us to the point that the crust-mantle boundary is a transition zone with a thickness of several kilometres.

At the Azores, eruptions carrying mantle and/or ultramafic cumulates are a rare exception and are linked with the movement of strike slip deep faults, while those containing olivine antecrysts are the rule. Fluid inclusion barometry in both cases provides similar depths (type I and type II), both at the scale of a single volcano (e.g. Pico) and at a regional scale (i.e., the islands of the Central Group). The depths resulting from the pressure recalculation are in good agreement with the deepest seismicity recorded during the last 22 years, defining a rather precise rheological/lithological boundary made up by rigid lithologies and in the absence of magmas.

The calculation of the depth of trapping of the fluid inclusions considered the ρ scheme reported in Zanon et al. (2020), developed for Pico volcano and applied to the highest density fluid inclusions (Table 2). This model integrated VP/VS vertical profiles (Matias et al. 2007) obtained from the detailed study of the seismic sequence associated with the ML5.8 earthquake of 1998 CE between Pico and Faial islands and integrated a statistically consistent dataset of fluid inclusions from Holocene eruptions from the central volcano. This scheme was applied to the volcanic systems at Faial, Graciosa and Terceira (Table 3) because of a similar age of volcanic history, age of the oceanic crust, values of the highest fluid inclusion density and deep hypocentres.

Table 3 Age, pressure, depth from fluid inclusions, and hypocentral depth
Full size table

As regards the volcanic systems of the other islands, there are no high-quality VP/VS seismic profiles available to produce local ρ schemes. Therefore, a simplified two-layer model (ρcrust = 2800 kg∙m−3 and ρmantle = 3215 kg∙m−3) has been used (Table 3). The oversimplifications associated with this model, however, seem to cause an overestimation in the calculation of the crust-mantle transition of the order of 2–3 km. In fact, the recalculation of literature data using the model developed for Pico volcano reduced by 2.2 km the depth of the crust-mantle transition below the eastern fissure zone of Pico (Zanon and Frezzotti 2013; Métrich et al. 2014) and by 1 and 0.6 km beneath the submarine and subaerial fissure zones at Terceira (Zanon and Pimentel 2015). A higher error is expected below volcanic systems located over a thicker crust, where there is a more complex stratigraphy. However, at S. Miguel and S. Jorge, there is significant seismicity to cross-check and validate the data from fluid inclusions, unlike Flores and Santa Maria, where seismicity is absent.

In agreement with the pressure scheme found, the deepest magma accumulation zones have been found at ~ 33 and ~ 31 km deep (± 0.02) below the oldest Santa Maria and Flores islands, located at the two opposite extremes of the region. As regards the more recent islands, S. Miguel, which is only 65 km far from Santa Maria, the deepest magma accumulation zones beneath the two fissure zones of the island are at ~ 29 km (± 0.04), while magma ponds at ~ 27 km (± 0.05) beneath Sete Cidades volcano, at 23 km (± 0.04) beneath Água de Pau, and only at ~ 20 km (± 0.09) deep beneath the extinct volcano of Nordeste to the east. The depths beneath the Faial-Pico system, which is set close to the MAR, are ~ 17.5–20 km (± 0.04). However, Graciosa and S. Jorge are equally distant from the MAR, and the deepest magma ponding zone is at a depth of 20.1 km and 26.3 km (± 0.04), respectively. This depth, beneath the submarine and the subaerial fissure zones at Terceira, reaches 20 and 17.9 km (± 0.04), respectively.

Fluid inclusions versus clinopyroxene-melt barometry

There is much literature on the magma system of oceanic volcanoes worldwide (see Barker et al. 2021, for a compilation). The works of the last 15 years rely upon the data provided by the use of the clinopyroxene-melt geobarometer in its original formulation (Putirka 2008) and subsequent recalibrations. In comparison with fluid inclusion barometry, this barometer provided overall deeper values than fluid inclusions at the Canary Islands and Fogo volcano (Hansteen et al. 1998; Hildner et al. 2011; Klügel et al. 2005; Stroncik et al. 2009). This would suggest that fluid inclusions are inadequate to define the full extent of the magma system of a volcano. However, new statistically robust studies of fluid inclusions on the recent eruptions of La Palma (Canary Islands) and Fogo (Cape Verde) extended the definition of the magma systems beyond the published limit attained by the clinopyroxene-melt geothermometer (Lo Forte et al. 2022; Zanon et al. 2022). As stated by Putirka (2008), the error of this barometer on calibrated data ranges between 150 and 200 MPa (i.e., a maximum error of about ± 7 km), while the average standard error on data not used for calibration is up to 450 MPa (Wieser et al. 2023). These pressures correspond to an estimated depth of ± 17.5 km.

Finally, the clinopyroxene population in a rock erupted in oceanic islands comprises xenocrysts, antecrysts, pheno and microphenocrysts, frequently showing complex zoning patterns. This large variety makes the interpretation of the data on the equilibrium condition very difficult.

Therefore, in comparison, this algorithm is less reliable than fluid inclusions, and a statistically robust database of fluid inclusions is sufficient to define, with good resolution, the full-scale magma system of a volcano.

The crust-mantle transition

All the information produced was used to draft the distribution of the deepest magma storage areas below the studied volcanic systems (Fig. 5). These deepest areas are interpreted to represent the crust-lithospheric mantle transition, following these lines of thought:

  1. i.

    The mafic magmas, which rapidly ascend through the crust, carried a heavy cargo of mafic minerals, which are in chemical disequilibrium with the host melt (i.e., they are antecrysts). These crystals show high-density, CO2-rich fluid inclusions whose trapping depths show coherency in the volcanic systems of the same sector of the archipelago and agree with the sources of the deepest tectonic seismic events recorded during monitoring activity in the last 22 years.

  2. ii.

    Fragments of mantle lithologies are quite rare in the Azores. The barometry from fluid inclusions in mantle xenoliths from Pico and Faial showed a deeper trapping area, not affected by seismicity, than the fluid inclusions in antecrysts in the carrier magmas (Zanon and Frezzotti 2013). A magma ascending through the lithosphere and with the capacity to hold a large amount of suspended mafic crystals should also contain mantle xenoliths, such as in the case of the Borgarhraun lava in Iceland (Mutch et al. 2019). Conversely, magma ponding at the crust-mantle boundary is sufficient to fractionate this heavy cargo.

  3. iii.

    Recorded deep seismicity has a pure tectonic signature. It is absent below fissure zones unless they activate during spreading events, which may interest the whole crust thickness (i.e., during 2022 seismicity at S. Jorge Island, scattered events were recorded even at a depth of 18 km, in good agreement with fluid inclusions data). More frequent deep seismicity is recorded at a transtensive fault (N45°) intersecting the west portion of Faial Island and the Dom João de Castro fault (N40°), which generates seismic events with magnitude values up to 3 and depths ≥ 20 km. Events occur frequently up to 15 km, in agreement with the rigidity of the upper crust, and then vanish with depth. At the MAR segment crossing the Azores Plateau, the recorded seismicity presents intermediate values, ranging from 15 to 20 km.

  4. iv.

    Two studies used the teleseismic wave receiver function to locate the crust-mantle boundary beneath the islands of the Azores. According to Spieker et al. (2018), this boundary is between 16 and 17 km deep, regardless of the distance from the MAR. Differently, Silveira et al. (2010), who applied a low-pass filter to seismic P-waves to suppress the signal noise, obtained a depth of 20 km beneath the islands closer to the MAR and 30 km beneath S. Miguel to the east. The dataset presented here encompasses the whole variability found at a regional scale by those two studies and even reveals accurate variability of the crust-mantle boundary at the scale of a single island.

Fig. 5
figure 5

a Oriented profiles showing the crust-mantle boundary variability across the islands of the Azores. Dashed lines mark the boundary either from microthermometric data or seismic swarms. Hypocentres projected on each profile occurred within a lateral distance of 15 km from both sides of the profile surface in a 3D projection, except for the events along the Mid-Atlantic Ridge, where a 100 km buffer was used. Earthquakes shallower than 2 km and isolated events have been filtered out. Depths resulting from microthermometry are represented by blue spots. b The a-a’’ profile runs from the Flores Island to Faial-Pico volcanic systems, passing through the Mid-Atlantic Ridge. c The b-b’ profile is across the island of S. Jorge and shows scattered ML3-4 seismicity. d The c–c’’ profile runs across the Graciosa-Terceira-S. Miguel islands and was split into two segments to fit in the figure. The c–c’ segment intersects the islands of Graciosa and Terceira and shows deep and intense seismicity WNW of Graciosa Island. The c’-c’’ segment goes from Terceira to S. Miguel islands and includes the Dom João de Castro Seamount (“DJC Smt” in the figure) and the Hirondelle Basin (“H Basin” in the figure). The lack of data between Terceira and the Dom João de Castro Seamount might be an artefact because of the lack of coverage by seismic stations. e The d-d’ profile intersects Pico, S. Jorge and Graciosa islands and, (f) the e-e’ profile goes from Santa Maria Island to the central volcanic system of Água de Pau in S. Miguel Island

Full size image

The crust-mantle boundary defined in Fig. 5 is represented by a dashed line in each profile, drawn from the value of either the deepest seismic events or the highest pressure resulting from the microthermometric study at each volcanic system. The main limitation of this model is the absence of information below oceanic basins between volcanic islands. As these basins were generated by recent tectonics after the formation of the plateau (Marques et al. 2013) and are a-magmatic, we assumed the minimum depth among the maximum values recorded by fluid inclusions at the regional scale (i.e., that of Pico volcano).

Overall, the Azores region is characterised by a crust thicker than 20 km, resulting from cycles of magma accumulation occurring since the first widespread fissural activity more than 39 Ma ago. The effect of magma accumulation below the islands is difficult to model for several reasons. (i) Three magmatic phases followed one another: the plateau phase, the volcanic elongated ridge phase, and the central volcano phase. Poor constraints on the beginning and end of each phase make it very difficult to estimate the volumes of involved magma. (ii) The shift from the elongated volcanic ridge to central volcano phases involved the shift from diffuse and fissural to focused magmatism beneath well-established conduit systems, defined by the periodical activation of N45-N60° faults. These latter faults play a major role in the withdrawal of mafic magmas from deep ponding areas (i.e., Zanon et al. 2020), generating local spatial heterogeneities in the distribution of mush layers and their thickness. (iii) Part of the interstitial magma (intracumulus) has been extracted from crystal mush, similarly to other volcanic islands (e.g., Mittal et al. 2022; Bouvier et al. 2022; Lissenberg et al. 2019). The ultimate consequence of this process acting throughout the crust is the settling of mush layers and the summit collapse of central volcanoes, as well as the late emission of intracaldera trachytes. (iv) Multiple cycles of magmatism contributed to a thick crust beneath the Flores and Santa Maria islands and beneath the central volcanoes in S. Miguel.

In the a-a’ profile (from Flores to Pico), the crust-mantle transition is traced at a depth range of 27.6–30.6 km. Flores Island is old and geophysically inactive, therefore, there is no seismicity to strengthen fluid inclusion data. The studied samples belong to the intermediate phase of activity (0.69–1.3 Ma). Eventual olivine-hosted fluid inclusions from the Holocene basalts (not found in the few samples collected) could provide even deeper ponding stages.

At the MAR, the source of the low-to-moderate energy seismicity (from ~ 3 to ~ 5 ML) is recorded at a maximum depth of ~ 15 km, while alongside the central volcano of Pico and the Faial Island, the cross-check of seismic data and fluid inclusion barometry provided a maximum depth of 17.7 and 20.7 (± 0.04) km. A main fault between the two islands generates deep seismicity up to a depth of ~ 20 km (Fig. 3). Magmatism at Faial started about 850 ka, interspersed by long periods of repose, erosion and tectonics (Hildenbrand et al. 2012). Pico volcano began to erupt ~ 53 ka (Costa et al. 2014) and rapidly developed a deep mushy magma system, from which olivine and clinopyroxene-rich lavas are extracted.

In profile b-b’, the crust-mantle transition below the S. Jorge volcanic systems is represented, along with the intense seismicity that started March 19 with events up to ~ 36 km deep. This island formed inside a graben (Marques et al. 2013), and the long-lasting magmatism (~ 600–700 ka for each cycle) produced a crust with a thickness that varies from 26.3 (± 0.04) km below the ~ 2 Ma old Topo fissure zone to 25.4 (± 0.04) km below the more recent Rosais-Manadas fissure zone. However, these depths can be overestimated by up to 3 km because of the application of the simplified stratigraphic model. The two parallel dashed lines represent the uncertainty because of the use of the simplified model. The difference in thickness along the island from WNW to ESE can be explained by the reduction in mantle melting degree, which occurred at the end of the early magmatism in the eastern third of the island (Hildenbrand et al. 2014), and by a different duration of magmatism.

Profiles c–c’ and c’-c’’ show the crust-mantle transition below the islands along the Terceira Rift (i.e., Graciosa, Terceira and S. Miguel, from WNW to ESE), the submarine volcano Dom João de Castro, and deep oceanic basins (e.g., Hirondelle Basins). The profile runs through a crust with ages ranging from ~ 15 Ma below Graciosa to ~ 40 Ma below S. Miguel (Müller et al. 2008). Overall, the crust-mantle transition depth increases from ~ 20 below Graciosa and offshore Terceira islands to ~ 30 km below the Dom João de Castro Seamount (only by seismicity) to S. Miguel and Santa Maria islands. Below Terceira Island, fluid inclusion barometry and seismicity data are not deeper than 17.6 km. As seismicity deepens again towards the SE, alongside the Terceira Rift, it is possible to hypothesise that the number of olivines analysed was not statistically sufficient to sample deeper ponding stages.

The profile along d-d’ crosses volcanic systems, which are at the same distance from the MAR (150–165 km), but with different ages and duration of magmatism (Table 3). Here is evidence of a relationship between the thickness of the crust and the age of the volcanic system. In this profile, it should also be considered the presence of Faial, which is roughly coeval to Graciosa and shows the same seismicity and fluid inclusion barometry.

Finally, the profile e-e’ runs from the old and eroded island of Santa Maria to the central volcanic system of Água de Pau at S. Miguel. Data are very similar, and all suggest a crust-mantle transition at ~ 30 km depth. Below the volcanic system of Água de Pau, a few measurements on fluid inclusions recorded 23.2 km. In this case, limited fluid inclusions data might have biased this result; in fact, at the two fissure zones nearby, fluid inclusions barometry provided values close to 30 km again.

Concluding remarks

Magmas that erupted at the Azores during the last 20 ka contain olivines and clinopyroxenes, which host CO2-rich fluid inclusions. High-density inclusions are commonly found as trails formed during re-equilibration events, whereas isolated inclusions are rare. An exhaustive microthermometry study combined with seismicity beneath the islands east of the MAR allows us to draw the following concluding remarks:

  • Barometric data from fluid inclusions and deep recorded seismicity located deep magma accumulation zones generally deeper than 20 km below the Azores region. Recorded earthquakes are tectonic, and no deep hybrid events have been recorded so far, thus excluding the presence of fluids and melts.

  • There is a very good agreement between the source of the deepest earthquakes and the sites of trapping inclusions in olivines.

  • Crust thickening in the region of the Azores is proportional to ageing and distance from the MAR, with thickness around 17–20 km below the islands closer to the MAR and younger than 1 Ma, around 30 km below the farthest island (S. Miguel), and those where magmatism lasted for more than 2 Ma (Santa Maria and Flores). S. Jorge Island, which is close to the MAR but has a magmatic history between 1 and 2 Ma, presents an intermediate depth.

  • Differences in the crust-mantle depth beneath various volcanic systems that are present on the same island are due to the interplay of cyclical magma accumulation with periods of erosion and tectonics. Only central volcanoes are intersected by faults with different orientations, which operate to contrast or enhance magma ascent. Tectonic quiescence promotes magma accumulation and evolution, with the formation of ultramafic cumulitic layers at depth. Fissure zones are not crossed by tectonic discontinuities other than the regional trend. Therefore, magma withdrawal is more effective, and the formation of deep cumulates is less important.

  • The cross-checked validation of seismicity and fluid inclusion barometry defines the deep roots of magma systems. This method should enhance a better detection of magma movement at depth during the early stages of a magmatic intrusion.