Time-Dependent Evolution And Source Heterogeneities of Ocean Island Basalts From a Weak Plume, São Jorge, Azores

Abstract

The geochemical composition of ocean island basalts (OIBs) from the Azores (AZ) reflects the spatial distribution, shape and temporal evolution of small-scale geochemical heterogeneities within their mantle plume source. Here, we investigate the time-related evolution of volcanism at São Jorge Island, Central AZ. New field observations, a magnetic survey, ⁴⁰Ar/³⁹Ar and ¹⁴C ages and geochemical data indicate that the fissural volcanic activity at São Jorge produced at least four main mafic volcanic complexes (V. C.). The oldest V. C., São João, produced the thickest lava piles at ca. 1.3 Ma in the eastern part of the island. After a period of quiescence, the Serra do Topo V. C. was produced at ca. 0.8–0.5 Ma in the central part of the island. The Rosais V. C. was emplaced between ca. 0.4 and 0.1 Ma on the entire island. Finally, the Holocene Manadas V.C. volcanism became active in the western part of the island and includes three historic eruptions (1580, 1808 and 1964 CE).

Magmas were formed at low melting degrees from a peridotitic mantle with possible minor contributions of recycled components. Olivine compositions and whole-rock trace element ratios discard a significant contribution from pyroxenitic source rocks. Melting temperatures (ca. 1420–1480 °C) were slightly higher than those of the ambient upper mantle. The four V.C. are characterized by distinct geochemical compositions in terms of incompatible trace elements and Sr-Nd-Pb isotopic ratios. The oldest V.C., São João, is characterized by Pb isotopic compositions (e.g. markedly negative Δ7/4 and 8/4 values) plotting well below the Northern Hemisphere Reference Line (NHRL). The Upper Pleistocene (Rosais V.C.) lavas from the north-western cliffs have compositions similar to enriched mantle (EM) end-member basalts (e.g. high ²⁰⁷Pb/²⁰⁴Pb at moderate ²⁰⁶Pb/²⁰⁴Pb; high Ba/Nb), which are rare among northern hemisphere OIBs. Finally, high ²⁰⁶Pb/²⁰⁴Pb (up to 20), reflecting contribution from a HIMU-type component characterizes the Holocene Manadas lavas and is occasionally found in lavas from other VCs from 1.3 to 0.1 Ma. These findings indicate that magmas from São Jorge and the nearby Central AZ islands were sourced from a strongly heterogeneous mantle plume, which displayed localized filaments of heterogeneous material that were rapidly exhausted (in ca. 0.2 Ma). The dominant component at São Jorge and in the Central AZ in general appears to be the HIMU-type end-member, which instead is not significant in the Eastern AZ. Possibly, the Central and Eastern AZ were produced by distinct branches of the AZ mantle plume, as would also be consistent with previous tomographic studies.

INTRODUCTION

Ocean Island Basalts (OIBs) provide fundamental information on the composition and evolution of the Earth’s mantle from which they derive. The generation of OIB magmas is mostly considered to be caused by mantle plumes rising from the deep mantle or from the 670 km discontinuity (Morgan, 1971; Manga et al., 1993; French & Romanowicz, 2014; Koppers et al., 2021). Worldwide OIBs display significantly different geochemical compositions resulting from mantle plume source heterogeneities (Zindler & Hart, 1986; Hofmann, 2003; Koppers et al., 2021; Weis et al., 2023). The spread of isotopic and trace element compositions in single islands or archipelagos depends on the heterogeneity of the mantle source, but is also controlled by the degree and depth of melting, and by the plume temperature and buoyancy (King & Adam, 2014; Hoggard et al., 2020; Weis et al., 2023). Ingeneral, enriched components (s.l.) can be preferentially extracted at low melting degrees in particular when they are present as fertile, low-solidus rocks in the mantle (such as pyroxenite or eclogite; Sobolev et al., 2007), while higher melting degrees tend to homogenize the composition of the erupted magmas, diluting the contribution of the enriched components (s.l.) in the erupted magmas (Stracke & Bourdon, 2009). Therefore, high-buoyancy and high-temperature plumes like Hawai’i or Iceland tend to produce erupted tholeiitic magmas with relatively minor geochemical variability when compared to colder plumes producing alkaline basaltic magmas (Weis et al., 2023). Numerical modelling studies (Farnetani et al., 2018) indicate that geochemical heterogeneities in mantle plumes may occur either as distinct filaments or blobs depending on the plume’s energy and on the viscosity contrast between the heterogeneity and the dominant plume material. In plumes with high buoyancy like Hawai’i or Tristan (King & Adam, 2014), distinct basalt compositions are observed in geographically separated alignments of volcanoes tapping compositionally distinct mantle filaments or stripes for periods of several million years (Abouchami et al., 2005; Weis et al., 2011; Farnetani et al., 2012; Hoernle et al., 2015). In archipelagos like the Canary or Pitcairn, the geochemical compositions of OIBs can progressively change over time when mantle regions (blobs) of distinct composition become dominant in the zone where mantle melting occurs (Taylor et al., 2020; Cordier et al., 2021). On the contrary, discrete and randomly distributed heterogeneous filaments are synchronously tapped by the erupted magmas for example at the Society Islands (Cordier et al., 2021).

Here, we investigate whether small-scale mantle heterogeneities are evident in the composition of basaltic magmas from São Jorge Island, Azores (AZ) archipelago (Portugal), and try to reconstruct their time-related variations (Fig. 1). Based on geochemical (Beier et al., 2007, 2010, 2022; Béguelin et al., 2017; Moreira et al., 2018) and geophysical data (Gente et al., 2003; Montelli et al., 2004, 2006; Silveira et al., 2006; Yang et al., 2006; Adam et al., 2013; Saki et al., 2015; O’Neill & Sigloch, 2018; Marignier et al., 2020; Bonatto et al., 2024), the magmatism of the AZ and its associated plateau might represent the surface expression of a mantle plume with moderately low buoyancy flux (0.38–0.85 Mg/s; King & Adam, 2014; Hoggard et al., 2020) and slightly elevated mantle potential temperatures (generally less than 1500 °C; Herzberg & Gazel, 2009). The AZ plume may be considered as a branch splitting at 670 km depth (mantle transition zone) from the large mantle plume giving rise to OIB magmatism in the Macaronesian archipelagos (Cape Verde, Canary, Madeira; Saki et al., 2015; Civiero et al., 2021). Branching of the AZ plumes has been as well imaged in the shallow mantle at ca. 200 km depth between the Central and Eastern AZ (Yang et al., 2006).

This study presents new geochemical and mineralogical data obtained on samples representing the main V.C.s that formed the subaerial portion of São Jorge Island. We also present a revision of the island’s volcano-stratigraphy, based on field relations, new ⁴⁰Ar/³⁹Ar and ¹⁴C ages, and magnetic data. This new dataset provides insights into the time-related evolution of the mantle source in the Central AZ. We also bring new constraints on the distribution, size and nature of mantle heterogeneities at play during recent pulses of the Azorean volcanism.

GEOLOGICAL SETTING

The AZ archipelago, in the Central Atlantic, lies on the triple junction between the North America, Eurasia and Nubia lithospheric plates (Adam et al., 2013; Miranda et al., 2014; Madeira et al., 2015, Fig. 1a and b). Plate margins are defined by the Mid-Atlantic ridge and by the Terceira Rift, a slow spreading transtensional structure between the Eurasian and the Nubian plates (Madeira & Ribeiro, 1990; Luis & Miranda, 2008; Hildenbrand et al., 2014; Storch et al., 2020; Beier et al., 2022). Formation of North-Northwest–South-Southeast-oriented horst and graben rift structures is associated with NNE–SSW horizontal shearing in the shallow mantle (Adam et al., 2013).

The nine Azorean islands form three geographic groups, Eastern, Central, Western AZ (Fig. 1b), which do not show any time-related migration of the magmatism (Larrea et al., 2018). Most Azorean islands are complex volcanoes, composed of both linear (fissural) and central volcanic systems. The products of linear volcanic systems are typically less evolved lavas (basalts to mugearites), while central volcanoes reach more evolved compositions (basalts to trachytes) in most cases related to caldera forming stages. The volcanic stratigraphy is not always easy to reconstruct on these islands, hindering the understanding of the time-related evolution of the volcanic products and magma compositions.

The Island of São Jorge, in the central group of the AZ islands, is anomalous in this archipelago since it was exclusively formed by fissural volcanic systems. The exposure to highly energetic marine conditions produced tall sea-cliffs whenever volcanism waned (Madeira, 1998; Madeira & Brum da Silveira, 2003) and created good outcrops of the older volcanic units. Furthermore, São Jorge is anomalous since, with Pico and Santa Maria, it lacks felsic rocks unlike the other Azorean islands (for a synthesis on the petrology of the AZ see Larrea et al., 2018).

According to previous studies (Forjaz & Fernandes, 1975; Hildenbrand et al., 2008; Marques et al., 2018), the oldest volcanic activity at São Jorge, the Topo Volcanic Complex (V.C.), occurred to the East of the Ribeira Seca fault (Fig. 1c), while more recent volcanism occurred in the West of the island (Rosais and Manadas V.C.). In the present study, we revise the previously proposed volcanostratigraphy, suggesting that the Topo V.C. should be divided into two V.C. with clearly distinct age and that thus at least four distinct V.C. were present on São Jorge. The youngest volcanic activity at São Jorge, the Manadas V.C., includes three post-settlement (mid-15th century) eruptions, the subaerial hawaiian/strombolian events of 1580 and 1808 CE and a submarine eruption in the year 1964 (Zanon & Viveiros, 2019). Both the 1580 and 1808 eruptions produced pyroclastic density currents (block and ash flows and wet surges, respectively) that caused some tens of casualties. Intense seismic activity in early year 2022 in the western part of the island was interpreted as resulting from an aborted eruptive event (Asensio et al., 2023).

SAMPLING

Sampling for geochemistry and geochronology (¹⁴C and ⁴⁰Ar/³⁹Ar) was mainly performed along sea-cliffs where the most complete volcanic sequences and related dykes are exposed. The two sections of São João and Vimes were sampled on the south coast of eastern São Jorge, while the north-eastern cliffs are mostly inaccessible for continuous sampling (Fig. 1c). On the western part of São Jorge, the Rosais V.C. was sampled along the trail that leads to the littoral platform of Fajã de João Dias (the toponym Fajã is the name given to littoral platforms at the base of sea cliffs produced by landslides or lava flows; Melo et al., 2018). The younger sequence of the Manadas V.C. is well exposed all over the ridge of the western part of the island. Its sampling was conducted to get a representative account of this volcanic phase, including the historical events. It must be noted that, while both the São João and João Dias sections sampled a single volcano-stratigraphic unit, the Vimes section covers three different units (referred to as São João, Serra do Topo and Rosais in the following text) as will be discussed later. Further details on the sampled sections are reported in the Electronic Appendix.

ANALYTICAL METHODS

⁴⁰Ar/³⁹Ar analyses (Table 1 and Electronic Appendix, Table S1) were obtained on groundmass samples with a MAP 215–50 mass spectrometer at Curtin University, Western Australian Argon Isotope Facility (WAAIF) of the John de Laeter Centre, following procedures outlined for example in Boscaini et al. (2022). Decay constant and standard ages are calculated after Renne et al. (2011).

Radiocarbon analyses were obtained on peat, wood, charcoal and paleosols underlying Manadas lava flows and pyroclastic deposits (Madeira & Brum da Silveira, 2003). About, 15 samples were analysed at Laboratório de Isótopos Ambientais (ITN/MCT, Sacavém, Portugal) following procedures for the ITN analyses are described in Madeira et al. (1995) and two additional samples at Beta Analytic (http://www.radiocarbon.com/about-carbon-dating.htm).

A land magnetic anomaly map was obtained using Mirone software (Luís, 2007) after removing the present magnetic field retrieved from IGRF 2010 (Finlay et al., 2010) for the whole island.

Olivine major element compositions were obtained by electron microprobe analysis (CAMECA SX50 at the CNR-IGG, Padova, Italy), while whole-rock major element and selected trace element contents were determined by X-ray fluorescence (XRF) at the University of Padova with a Philips PW2400 spectrometer (procedures in Boscaini et al. (2022). Trace elements were also analysed by inductively coupled plasma–mass spectrometry (ICP-MS) at the University of Bretagne Occidentale at Brest (France), following analytical protocols described in Barrat et al. (1996).

Radiogenic isotope ratios of Sr, Nd and Pb were measured at the Department of Earth Sciences (University of Geneva, Switzerland) using a Thermo Neptune PLUS Multi-Collector ICP-MS. The method is described in detail in Chiaradia et al. (2020) and Béguelin et al. (2015). The details of the analytical methods can be found in the Electronic Appendix.

RESULTS

⁴⁰Ar/³⁹Ar ages

Six whole-rock ⁴⁰Ar/³⁹Ar ages were obtained in this study (Fig. 2; Table 1). The analysed samples correspond to lava flows and were collected at the São João, Vimes (Lower and Middle Vimes) and João Dias sections. Five rocks yielded largely concordant plateau ages, defined by more than 95% of the total emitted gas, with MSWD close to 0.5 and probability close to 0.9. Four of these plateau ages (samples AJ7, AJ14, AJ71 and AJ30) are indistinguishable from the respective inverse isochron ages (³⁹Ar/⁴⁰Ar vs. ³⁶Ar/⁴⁰Ar; not shown), which yield initial ⁴⁰Ar/³⁶Ar overlapping with the atmospheric value (298.56). The other plateau age (sample AJ32) shows a slightly saddle shaped spectrum, possibly suggesting the presence of excess argon, i.e. an apparent age older than the magmatic crystallization age. For this sample, we retain the isochron age, which is statistically robust (MSWD 0.7, probability 0.8) and yields an initial ⁴⁰Ar/³⁶Ar of 362 ± 20, confirming the presence of excess argon. The sixth sample (AJ66) did not yield a plateau age, yet the inverse isochron is statistically robust (MSWD 0.5, probability 0.4) and its initial ⁴⁰Ar/³⁶Ar (280 ± 6) is close to the atmospheric value.

The retained ages confirm the presence of an early volcanic activity at 1.36 ± 0.17 Ma to 1.35 ± 0.16 Ma for the São João section (samples AJ7 and AJ14) and for the lower part of the Vimes section (AJ66, inverse isochron age 1.37 ± 0.18 Ma). A plateau age of 722 ± 25 ka was obtained for sample AJ71, collected in the middle part of the Vimes section. Samples from the João Dias section constrain the age of the Rosais lava flows (samples AJ30 and AJ32: 407 ± 54 and 256 ± 44 ka, the latter being an inverse isochron age).

The new ⁴⁰Ar/³⁹Ar and previously published K-Ar and ⁴⁰Ar/³⁹Ar age data (Féraud et al., 1980; Hildenbrand et al., 2008; Ribeiro, 2011; Marques et al., 2018) define the ages and duration of the subaerial volcanic activity at São Jorge (Table 2; Figs 2 and 3, all available K-Ar and Ar/Ar ages are also reported on the map of Fig. 1c). The oldest age (1.85 ± 0.04 Ma) was found by Marques et al. (2018) for a lava flow cropping out at the base of the north-eastern cliff (Fig. 1c). Volcanic activity ranging from 1.70 ± 0.02 to 1.21 ± 0.02 Ma formed the lava sequence exposed at the south-eastern cliffs (São João and lower Vimes sections). Ages spanning from 811 ± 17 to 543 ± 4 ka characterize the second oldest unit in the eastern portion of the island at Serra do Topo and on the southern and northern sea cliffs immediately east of the Ribeira Seca fault (Féraud et al., 1980; Hildenbrand et al., 2008; Ribeiro, 2011 and present study, Fig. 1C). Ar/Ar ages of 407 ± 54 and 256 ± 44 ka were obtained for lava flows from the João Dias section. They overlap with previous ages for lava flows from the easternmost tip (near Topo village; Féraud et al., 1980) and the western part of the Island (near Velas village, Hildenbrand et al., 2008 and at João Dias, Ribeiro, 2011). Slightly younger ages ranging from 140 ± 10 to 110 ± 70 ka were obtained by Féraud et al. (1980) and Ribeiro (2011) to the east of Velas (see Electronic Appendix Fig. S1) and near Ribeira Seca in the central-eastern part of the islands.

Considering these combined K/Ar and Ar/Ar age data, the Pleistocene volcanic activity on São Jorge can be divided into three main V.C.s (Figs 1 and 3): the ca. 1.8–1.2 Ma old unit (hereafter São João V.C.); the ca. 0.8–0.54 Ma old unit (Serra do Topo V.C.); and the ca. 0.4–0.1 Ma old unit (Rosais V.C.). A period of quiescence may have occurred between ca. 1.2 and 0.8 Ma, (i.e. between São João and Serra do Topo V.C.), while no major volcanic gap is apparent between the end of the Serra do Topo and the onset of the Rosais V.C.

Radiocarbon ages

New and previously published (Madeira & Brum da Silveira, 2003) radiocarbon ages obtained for the Manadas V.C. indicate a Holocene age for the younger volcanic phase in São Jorge, consistent with field observations. The 21 conventional ¹⁴C ages (Table 3) range from 7170 ± 40 to 150 ± 40 years BP (BP = before 1950 CE). These ¹⁴C ages allowed dating at least 11 eruptions, including the three post-settlement eruptions (details on the sampling and dated materials are reported in the Electronic Appendix). Some ages overlap within error and may date the same event. For example, the humic acid extract from the paleosoil underlying a lapilli fall layer on the east flank of Pico da Esperança (SJ10) yielded the same age as Pico dos Loiros eruption (SJR14; Table 3). The age distribution indicates intervals of 100 to 750 years between eruptions since 3740 BP, although many events from Manadas V.C. remain undated.

Magnetic anomalies

At the island scale, magnetic anomalies vary between 4000 and −4000 μT (Fig. 4) with a dominance of positive ones. Such high amplitude variation and spatial predominance of positive anomalies is in good agreement with the known magnetization for the island of São Jorge from an aeromagnetic survey (Miranda et al., 1991) and a paleomagnetic study (Silva et al., 2012). The predominance of positive anomalies indicates that a significant volume of lava flows erupted during the Brunhes chron, i.e. during the last 0.78 Ma (Quidelleur et al., 2003). Nonetheless, high intensity negative anomalies are also recognized on the eastern and central eastern regions of the island (Fig. 4). This implies the presence of a basement formed by volcanic sequences with reverse polarity acquired during Matuyama chron (0.79 to 2.59 Ma). Most of the island, however, presents low to moderate anomalies (mostly between −600 and 600 nT). The subdued anomalies in areas presenting volcanic sequences of Rosais and Manadas V.C.s (<0.5 Ma; see above) indicate that the actual subaerial volcanic building started to be constructed during Matuyama (i.e. São Jorge Island started as a Matuyama proto-island that is underlying Brunhes volcanism). This interpretation is supported by the ages (Hildenbrand et al., 2008; Silva et al., 2012; Marques et al., 2018; this study) obtained for the older volcanic sequences (São João V.C. and lower part of Serra do Topo V.C.) that crop out in the eastern part of the island and coincide with high amplitude negative anomalies.

Classification and petrography of the rocks

In the following sections, we group the analysed samples as follows, based on the geochronological data presented above and on field evidence (Figs 1C and 3: 1) the São João V.C. (ca. 1.4–1.2 Ma) including all 14 samples from the São João sequence and the 10 samples from the lower part of the Vimes section (São João-Lower Vimes), up to sample AJ66, dated at 1.37 ± 0.18 Ma, and its immediately adjacent samples AJ67 and AJ68; 2) the Serra do Topo V.C. (ca. 0.8–0.6 Ma), which includes four samples from the middle part of the Vimes section, immediately adjacent to sample AJ71 (Middle Vimes); 3) the Rosais V.C. (ca. 0.4 to 0.1 Ma) formed by the 4 topmost samples from the Vimes section (Rosais-Upper Vimes) and all 13 samples from the João Dias section (Rosais-João Dias); 4) the Holocene Manadas V.C. (ca. <7200 years BP), recognized based on field and geomorphological criteria and radiocarbon ages.

The analysed rocks (Table 4, the complete data set is reported in the Electronic Appendix, Table S2) classify as basalts (30), trachybasalts (23, 21 of which can be defined as hawaiites, being sodic in composition) and mugearites (5) according to the total-alkali-silica classification (Le Bas et al., 1986; Fig. 5). Most basalts are moderately nepheline normative (up to 6%) and can therefore be classified as alkali basalts. However, four basalts (all from the São João section, São João V.C.) are olivine/hypersthene normative. Most Rosais samples (all except one from João Dias and one from Upper Vimes) and one Manadas sample are potassic (wt.% Na₂O-K₂O < 2.0), whereas all the other samples are sodic (Na₂O-K₂O ≥ 2.0).

The sampled rocks exhibit aphanitic groundmasses. They range from nearly aphyric (rare) to moderately or highly porphyritic (see the Electronic Appendix, Fig. S2). The rocks are frequently vesicular and contain phenocrysts and microphenocrysts and sometimes glomeroporphyritic aggregates in a fluidal groundmass. In a few samples, phenocrysts and microphenocrysts are more abundant than the groundmass (i.e. these rocks have cumulitic textures).

Mineral assemblages consist essentially of olivine (some with oxide inclusions), clinopyroxene and plagioclase occurring as phenocrysts, microphenocrysts and groundmass crystals. The groundmass is primarily made up of plagioclase laths, Fe-Ti oxides, rare apatite and a cryptocrystalline felsic interstitial filling. The crystallization sequence can be reconstructed in a few samples where olivine always crystallizes before clinopyroxene, which is later followed by plagioclase, as typical of alkaline basalts.

Despite similar textural features and similar mineralogy observed in all the sampled sections, the variability of mineral modal abundance expresses inter-section variations and differences among the V.C. Samples from the São João V.C. are characterized by prevalence of optically zoned plagioclase phenocrysts over rare olivine and clinopyroxene. Several São João rocks are strongly porphyritic or cumulitic (up to ca. 30 vol.% represented by large plagioclase crystals), but a few aphyric lava flows are present at the base and towards the top of the section (samples AJ01 and AJ11). Samples from the Rosais V.C. (João Dias and Upper Vimes sections) show large proportions of olivine and clinopyroxene relative to plagioclase. Manadas samples contain several large, zoned plagioclase crystals and less abundant clinopyroxene and olivine, and are generally quite porphyritic. A few Manadas samples (AJ51, AJ53, AJ77) yield amphibole with variable degrees of resorption rims. On the contrary, amphibole was not found in the Pleistocene lavas.

Mineral chemistry

Olivine crystals (26 crystals in total) from all V.C. at São Jorge were analysed (the data are reported in the Electronic Appendix, Table S3). Forsterite (Fo) contents in the crystal cores vary between Fo₈₇ and Fo₇₃ and in general decrease at the outermost rims to Fo₇₀ or less (Fig. 6). The Fo content of cores indicates that they crystallized from melts with distinct degrees of evolution, ranging from Mg# 67 to 47, respectively (assuming an olivine/liquid K_D(Fe/Mg) = 0.30 +/−0.03; Roeder & Emslie, 1970). All analysed olivine crystals yield low NiO contents (<0.30 wt.%, Ni < 2400 ppm; Electronic Appendix, Fig. S3), whereas CaO contents vary between 0.20 and 0.30 wt.% for most analyzed samples. This indicates that they are not high-pressure xenocrysts (Hirschmann & Ghiorso, 1994) or xenocrysts from peridotitic nodules, which usually yield low CaO (<0.10 wt.%).

Olivine cores of Manadas samples are in equilibrium with the host whole-rock (for a mineral/rock K_D(Fe/Mg) = 0.30 +/−0.03; Roeder & Emslie, 1970; Fig. 6a; Electronic Appendix, Table S3). On the contrary, most Rosais V.C. rocks yield olivine cores with lower Fo than expected for equilibrium conditions while about half São João rocks yield olivine with higher Fo than equilibrium conditions. This indicates multiple causes for olivine-liquid disequilibrium, at least when considering the whole rock as a proxy of the magma composition.

Geochemical compositions of whole-rock

Major and trace elements

The analysed whole-rock compositions (Fig. 7; Table 4) range from little evolved, with MgO up to 10.4 wt.% (and Mg# up to 67), to fairly evolved compositions (MgO = 3.5 wt.%, Mg# = 40; Mg# is defined as 100*Mg/(Mg+Fe²⁺), with Fe³⁺/Fe²⁺ = 0.18, consistent with moderately oxidizing conditions typical of alkaline magmas and with the constraints obtained from MELTS modelling (see below). The Loss On Ignition of all samples is lower than 1.6 and < 1.0 wt.% in most samples, indicating that alteration is negligible.

In general, SiO₂, Al₂O₃, Na₂O, K₂O and P₂O₅ increase at decreasing MgO and CaO, while TiO₂ and Fe₂O_3tot are near constant in high MgO samples and then decrease in more evolved samples (MgO < 5 wt.%). Major element compositions define significant differences among the volcanic units. Compared to the other complexes, rocks from Manadas, the youngest V.C., are relatively enriched in Na₂O, P₂O₅ and depleted in SiO_2, at a given MgO. Rosais rocks from both the João Dias and Upper Vimes sections are characterized by generally high K₂O and CaO/Al₂O₃ and tend to be fairly MgO-rich (5.4–10.4 wt.%). High P₂O₅ and low MgO characterize Serra do Topo (Middle Vimes) samples, while São João V.C. rocks (from the São João and Lower Vimes sections) tend to be the most evolved (i.e. Mg-poor) and Si-rich. They are depleted in Na₂O, K₂O, P₂O₅ and Al₂O₃, except for two Al- and Ca-rich samples, which are anomalously rich in plagioclase crystals (ca. 25 and 30% modal). Only the São João V.C. sample collected at the base of the exposed Vimes sequence is significantly less evolved (AJ59, MgO = 7.97 wt.%) when compared to other São João V.C. samples.

Systematic space- and time-related differences are also observed in terms of incompatible trace elements for São Jorge basalts (Fig. 8). Manadas rocks, Rosais-Upper Vimes and the two topmost samples from São João-Lower Vimes show similar trace element compositions, e.g. higher Nb/La (>1.60) relative to the other groups. Rosais V.C. rocks from the João Dias section have higher Ba, Rb and light rare earth elements (LREE) and slightly higher Ba/La, Ba/U, Ba/Nb, La/Yb, Nb/Zr and lower U/La and Nb/La compared to rocks from the other units at similar MgO, including the coeval Rosais-Upper Vimes flows. São João V.C. samples are amongst the most enriched in incompatible trace element contents, which is consistent with their generally evolved composition. However, they are characterized by low Nb/Zr, Ba/La and La/Yb.

On multi-element diagrams (spidergrams; Fig. 9), general features for all samples include light vs heavy REE enrichment, a positive Nb anomaly and a moderate enrichment of large ion lithophile elements (LILE) compared to LREE. São João and Manadas samples show a moderate negative K anomaly (i.e. mantle normalized K_N < Nb_N and < U_N; (K/U)_N = 0.6–0.9), while in Rosais V.C. samples from the João Dias section (but not in Rosais-Upper Vimes) the negative K and positive Nb anomalies are generally less pronounced or absent (K/U_N = 0.8–1.3; Nb/Nb* = Nb_N/(Th_N x La_N)^1/2 = 1.5–1.2). Ti shows a slight positive anomaly for Rosais and some Manadas samples and a negative anomaly for Serra do Topo and most São João V.C. lavas. Serra do Topo V.C. lavas display trace element patterns different from those of the São João-Lower Vimes and Rosais-Upper Vimes lavas. Serra do Topo samples are characterized by high trace element contents, by a positive P anomaly, and by moderately negative Sr, Zr-Hf and Ti anomalies (see also the Electronic Appendix, Fig. S4).

Sr-Nd-Pb isotopic compositions

Fifteen samples from all four V.C. were selected for Sr-Nd-Pb isotopic analyses, although we focused most of our analyses on the São João, Vimes and João Dias sequences, which were not previously analysed The obtained isotopic data display a relatively large spectrum of compositions (e.g.  ⁸⁷Sr/⁸⁶Sr = 0.7034–0.7042; ¹⁴³Nd/¹⁴⁴Nd = 0.5130–0.5128; ²⁰⁶Pb/²⁰⁴Pb = 18.98–20.13; Table 3), which are distinct for the four V.C.s (Fig. 10). As for incompatible trace element compositions, space- and time-related differences are observed for the isotopic compositions. Three end-member compositions can be defined: 1) moderately low ¹⁴³Nd/¹⁴⁴Nd (ca. 0.51295) and ⁸⁷Sr/⁸⁶Sr (ca. 0.7035) and quite high Pb isotopic compositions (²⁰⁶Pb/²⁰⁴Pb > 19.8) are shown by the two Manadas lava flows and by one São João-Lower Vimes, one Serra do Topo and one Rosais-Upper Vimes basalt; 2) Rosais V.C. samples from the João Dias sequence display high ⁸⁷Sr/⁸⁶Sr and low ¹⁴³Nd/¹⁴⁴Nd (<0.51290), while their Pb isotopes are characterized by low ²⁰⁶Pb/²⁰⁴Pb (ca. 19.0) at high ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb, i.e. positive Δ7/4 and Δ8/4 (Δ values expressed as vertical deviation from the NHRL; Hart, 1984); 3) all São João V.C. rocks from the São João sequence and the oldest São João-Lower Vimes sample are characterized by high ¹⁴³Nd/¹⁴⁴Nd (ca. 0.51296–0.51302) and intermediate Pb isotopic compositions (e.g.  ²⁰⁶Pb/²⁰⁴Pb = 19.28–19.53) plotting below the NHRL, i.e. at negative Δ7/4 and Δ8/4.

Sr-Nd-Pb isotopic ratios are in general correlated with variations of trace element ratios such as La/Yb and Nb/La (Fig. 11). Rosais samples from the João Dias section plot at extreme values indicating combined enrichment of some isotopic compositions (including high ⁸⁷Sr/⁸⁶Sr, Δ7/4 and Δ8/4, but not ²⁰⁶Pb/²⁰⁴Pb) and of La/Yb and Ba/Nb and decrease of the positive Nb anomaly (expressed by Nb/Nb*; Fig. 11e and f). São João samples show in general relatively depleted Sr-Nd isotopic compositions and low La/Yb (Fig. 11), and their Ba/Nb is inversely correlated with their Nd isotopic composition. Finally, the topmost São João-Lower Vimes sample, Serra do Topo, Rosais-Upper Vimes and Manadas samples have high ²⁰⁶Pb/²⁰⁴Pb combined with high Nb/La and Nb/Nb^* (Fig. 11d-f).

DISCUSSION

Volcano-stratigraphy and geochronology

The volcano-stratigraphy presented in this work is a revision of former stratigraphic schemes (Forjaz & Fernandes, 1975; Hildenbrand et al., 2008; Marques et al., 2018). Our conclusions are based on a review of all new and previously published geochronological data (Figs 1c and3), on a magnetic survey (Fig. 4), stratigraphic and geomorphologic field observations, and stereoscopic aerial photo interpretation. The main changes to the former volcano-stratigraphic schemes relate to the eastern half of the island. In this part of the island, previous studies considered a single volcanic unit (Topo V.C.), while we now propose that it includes at least three volcano-stratigraphic units.

The studies of Hildenbrand et al. (2008) and Marques et al. (2018) revealed the presence of an older volcanic phase dated at ca. 1.85–1.2 Ma. This oldest event indicates that subaerial volcanism on São Jorge started around the Pleistocene Gelasian-Calabrian boundary (Marques et al., 2018). Our new age determinations show that the ca. 1.3 Ma unit is present not only at the São João section (Fig. 1c) but also in the lower part of the Vimes section, thus having a larger area than previously thought. This is also consistent with the strong negative magnetic anomalies in both localities (Fig. 4), indicative of a Matuyama age (0.79 to 2.56 Ma). We name this unit the São João V.C. after the most accessible and best exposed location.

Two further volcanic events occurred on the eastern half of the island. Remnants of the activity at ca. 0.8–0.5 Ma were recognized on the central northern cliffs (Fajã dos Cubres and Fajã de Santo Cristo; Hildenbrand et al., 2008; Ribeiro, 2011; Marques et al., 2018) and on the central southern cliffs (middle part of the Vimes section and between Vimes and Fajã de São João; Hildenbrand et al., 2008; this study). We propose to name this unit Serra do Topo V.C. The contact between this and the underlying unit of the São João V.C. was observed from sea on the cliffs. This contact plunges to the west and disappears at sea level just east of Fajã dos Vimes. The São João-Lower Vimes samples are exposed due to fluvial incision of the south cliff forming a localized window.

The upper Pleistocene Rosais V.C. (ca. 0.4–0.1 Ma) is mainly exposed in the north-western part of the island at the João Dias section, as well as west of Fajã do Ouvidor, at Velas Bay and near Rosais village. Rosais lavas were dated at ca. 0.4–0.1 Ma (Hildenbrand et al., 2008; Ribeiro, 2011; this study; Fig. 1c). We propose that the Rosais V.C. is represented not only in the western part of the island, as suggested by previous studies, but also in the eastern part (Fig. 1c). Major element compositions of rock samples from the Upper Vimes section are significantly different from those of the underlying rocks dated at ca. 0.7 and 1.3 Ma (Serra do Topo and São João V.C.; Electronic Appendix, Fig. S4), but they are similar to those of the least evolved Rosais rocks from the João Dias section (Figs 7 and 8) and to a sample collected within this same area by Millet et al. (2009). Rosais rocks may also be present in the easternmost part of the island. The easternmost tip of São Jorge (area between the villages of Santo Antão and Topo; Fig. 1c and Electronic Appendix Fig. S1) displays a set of cinder cones and related lava flows with a degree of degradation like that observed in Rosais cones in the western part of São Jorge. This would be supported by the 110 ± 70, 140 ± 50 and 280 ± 90 ka ages presented by Féraud et al. (1980) for the eastern part of the island (Fig. 1c). The presence of Rosais V.C. in the eastern region of São Jorge is also supported by a strong positive magnetic anomaly in agreement with a Brunhes age for that unit (Fig. 4). Thus, the eastern half of São Jorge Island is formed by products of all the volcanic units recognized so far at São Jorge except for the Manadas V.C. The simplified geological map of São Jorge in Fig. 1c incorporates all this information.

In summary, the subaerially exposed structure of São Jorge was built during at least four main eruptive phases. Considering geochronology, magnetic and field evidence, we conclude that the oldest exposed building phase in São Jorge is mainly represented by the ca. 1.3–1.2 Ma São João V.C. and could include the 1.85 Ma old flows from the northern cliffs (Marques et al., 2018). This eruptive phase may thus have lasted for ca. 0.6 My and was followed by a period of volcanic quiescence of about 0.4 My. The following phase, Serra do Topo V.C. lasted ca. ~0.3 My and includes the Middle Vimes lava flows and coeval flows from the northern cliffs (Fig. 1c). This volcanic phase was followed by a short period of quiescence that may have lasted <0.1 My. The Rosais building phase lasted for ~0.3 My and included volcanic products erupted on the western part of the island and probably also on its eastern half (Fig. 1c). A time lapse of ~100 ky separates the volcanism of Rosais from the Manadas volcanic phase, whose products are found only west of the Ribeira Seca fault. Manadas volcanism is Holocene in age and possibly younger than 8000 years (Table 3). This Holocene volcanism is coeval with the recent eruptive activity in west Faial (Di Chiara et al., 2014) and Pico (Nunes, 1999), suggesting a phase of increased magmatism in this area of the archipelago during the Holocene.

Closed- and open-system differentiation

Closed-system fractional crystallization processes were modelled with MELTS (Ghiorso & Sack, 1995). The MELTS runs (Fig. 7) were performed starting from three relatively high-MgO parental magmas (AJ59, AJ32, AJ80; São João (Lower Vimes), Rosais and Manadas V.C., respectively). Liquid lines of descent (LLD) reported in Fig. 7 were obtained for a pressure of 0.6 GPa (as consistent with average pressure estimates in Zanon et al., 2023), FMQ-buffered oxygen fugacity and near dry to moderately hydrous starting compositions (0.1 and 1.0 wt.% H₂O). LLDs calculated for low pressure (e.g. 0.2 GPa; not shown) are significantly different from observed compositions.

Fractional crystallization modelled with MELTS shows that early crystallizing minerals are always olivine and clinopyroxene, the latter being slightly favoured at higher pressure and water content. Plagioclase and oxides join the crystallizing assemblage more than 150 °C after saturation of the mafic silicates. Only for São João samples, which are quite evolved, plagioclase saturation occurs relatively early (ca. 50 °C after clinopyroxene) for poorly hydrous melt compositions.

The calculated crystallization sequence is consistent with petrographic observations and with the moderately alkaline character of São Jorge magmas. However, in general, the overlap between modelled LLDs and observed compositions is not perfect (e.g. for SiO₂, A₂O₃, Fe₂O₃tot; Fig. 7). Even lavas from a single V.C. cannot all be derived from a single (common) parental magma, but instead require distinct mantle-derived primary magmas, as shown for example by the Na₂O vs. MgO diagram (Fig. 7). The mismatch between calculated closed-system LLDs and observed compositions also suggests that open-system differentiation may have been important. This is consistent with reverse zoning in some olivine crystals as well as with the observed disequilibrium between olivine and whole-rock compositions (Fig. 6). Similar interpretations were proposed for other OIBs, which frequently show evidence of open system processes in the deep oceanic crust (Ubide et al., 2021) or possibly in the lithospheric mantle (Zanon et al., 2023).

Differentiation of OIBs in the oceanic crust and lithospheric mantle may be associated with contamination processes. For example, the enriched isotopic composition (positive Δ7/4 and Δ8/4) of Rosais basalts (Figs 10 and 11) may result from assimilation of local marine sediments of terrigenous origin. However, this hypothesis is not consistent with: 1) the lack of correlation between major element compositions (e.g. MgO, SiO₂) and the isotopic values (e.g. with Δ7/4); 2) the fact that São Jorge Island rests on a horst, while sedimentation is probably more developed in the nearby Terceira graben (cf. Marques et al., 2018); 3) the far distance of the AZ from the continents, which rules out a continental origin for the local marine sediments. Rather, sediments in this region are expected to be dominated by eroded volcanics, ash and hydrothermal alteration (Schmidt et al., 2020) and thus would have compositions different from typical continental sediments. Moreover, it would be difficult to explain why assimilation of sediments would be limited only to the Rosais magmas from the João Dias sequence.

Another possible contaminant is represented by the MORB layer of the underlying oceanic crust with depleted isotopic signature. Based on oxygen and radiogenic isotope evidence, previous studies have shown evidence for interaction of AZ magmas with the local, underlying oceanic crust or lithosphere (Millet et al., 2009; Genske et al., 2012). However, Rosais samples have trace element contents and ratios and Sr-Nd-Pb isotopic compositions (Fig. 10) that are clearly different from present-day MORB from the nearby Mid-Atlantic Ridge (Dosso et al., 1999) arguing against a significant contamination by this local shallow component as an explanation for the composition of Rosais magmas.

Mantle melting

Mantle potential temperatures have been calculated after Putirka (2008) for basalt compositions having MgO > 8.8 wt.%. To obtain the primary magma composition (15 wt.% MgO), 12%–17% olivine has been added to the basaltic compositions. Petrographic observations, MELTS modelling and calculated crystallization temperature suggest that olivine crystallization precedes clinopyroxene saturation in high-MgO basalts. However, we cannot firmly rule out that some amount of clinopyroxene and olivine co-crystallized along the cotectic in magmas with ca. 9–10 wt.% MgO. Therefore, the calculated mantle potential temperatures should be considered with caution and are maximum values.

Considering a 1 wt.% H₂O content of the primary basalt (consistent with Beier et al., 2012), the calculated mantle potential temperatures range from 1450 to 1520 °C for a pressure of 2.8–3.2 GPa. The Lee et al. (2009) geobarometer yields slightly lower values for the least evolved São Jorge basalts, i.e. 2.5 ± 0.1 GPa and 1430 ± 10 °C. These calculated mantle potential values are consistent with those calculated by Herzberg & Gazel (2009) for the AZ (ca. 1430–1470 °C) and with those calculated by Beier et al. (2012) for São Jorge (ca. 1490–1540 °C) and for the AZ in general (ca. 1430–1540 °C). Overall, the obtained mantle potential temperature for São Jorge basalts is slightly higher than the typical ambient mantle potential temperatures extrapolated from MORB (ca. 1400 °C; Putirka, 2008; Herzberg & Gazel, 2009; Lee et al., 2009) suggesting that São Jorge basalts are issued from a moderately hot mantle as has been proposed, among others, by Putirka (2008) and Herzberg & Gazel (2009) for the AZ.

Olivine major and minor element compositions (Electronic Appendix, Table S3) can be used to infer the presence of pyroxenitic vs peridotitic source rocks (Sobolev et al., 2007). When compared to OIBs for which a pyroxenitic source component has been inferred (Sobolev et al., 2007; Delavault et al., 2015), olivines from São Jorge and other Central AZ basalts (Beier et al., 2012; Larrea et al., 2018 and references therein) generally are low in Ni (<2500 ppm; 100xNi/Mg < 1.2 for Fo > 80 olivines; Electronic Appendix, Fig. S3) arguing against a significant contribution from pyroxenitic source rocks. This conclusion should be considered cautiously as AZ olivines were not analysed following analytical protocols optimised for minor element quantification (as in, Delavault et al., 2015) and because high-Fo olivines are rare. Nonetheless, the general Fo vs Ni or vs. Mn/Fe variations of the Central AZ olivines (including the here studied samples) point to compositions for Fo-rich olivines that are clearly different from those inferred to indicate a pyroxenitic source component (Delavault et al., 2015; Sobolev et al., 2017). Consistently, the roughly negative correlation of Sr-Nd-Pb isotopic ratios with ²³⁰Th and ²³¹Pa excess (Bourdon et al., 2005; Prytulak & Elliott, 2009) does not support a significant presence of pyroxenitic lithologies in the mantle source of Central AZ basalts.

The variability observed in trace elements and Sr-Nd-Pb isotopic compositions of São Jorge rocks (Figs 8-11), as well as LLDs calculated with MELTS (Fig. 7) suggest that the different volcanic units from São Jorge require distinct parental magmas. To constrain mantle melting processes, we used the REEBOX-PRO code (Brown & Lesher, 2016) which models incompatible trace element variations during polybaric decompressional mantle melting. As none of the studied samples can be envisaged as representative of a primary mantle melt, it should be considered that ca. 20% fractional crystallization of olivine and clinopyroxene from a primitive basalt would slightly increase incompatible trace element ratios such as La/Yb and La/Gd (by <10%) and Ba/Nb, Nb/La, Gd/Yb (by <3%, considering fractionation of equal amounts of olivine and clinopyroxene and partition coefficients of Bédard (2014).

Calculated compositions of pooled melts are shown in Fig. 12, while pressure vs. temperature, melting degree variations and the conditions for which calculated and observed compositions are closest, as well as all considered mantle compositions are reported in the Electronic Appendix (Fig. S5 and Table S4). The reported melting curves (Fig. 12) were calculated for mantle potential temperature (Tp) ranging from 1420 to 1480 °C. Models produced at higher (>1500 °C) and lower (<1400 °C) mantle potential temperatures (not shown) resulted in more extreme calculated compositions that did not fit the observed values. The starting mantle source composition is an undepleted peridotite in terms of REE contents (i.e. primitive mantle REE values; Sun & McDonough, 1989), either dry or slightly hydrous (100 ppm of H₂O). Moreover, we tested a mixed mantle source with 98% peridotite and 2% pyroxenite (black continuous curve; G2 pyroxenite from Pertermann & Hirschmann, 2003; Spandler et al., 2008) or a peridotite enriched by 0.4% of continental crust (dashed black curve; upper crust composition from Rudnick & Gao, 2003). In the latter case, the enrichment is purely geochemical, following the approach of Boscaini et al. (2022), and the starting lithology is thus pure peridotite. A pre-existing lithospheric thickness of 50 km is assumed (Spieker et al., 2018), with a spreading rate of 0.5 cm/year (consistent with the low spreading rate of the Terceira Rift).

At such conditions, melting would start at ca. 3.6–3.8 GPa for the peridotite (at >4 GPa for the pyroxenite) and extraction of pooled melts similar to São Jorge basalts should occur at ca. 2.7–3.1 GPa and ca. 2%–3% melting (cumulative amount of pooled melts) of a peridotite with Tp of 1420–1480 °C. Calculated melts produced from a peridotitic source provide a relatively good fit with observed compositions, while those produced from the pyroxenite do not overlap with observed compositions (Fig. 12). The absence of pyroxenite in the mantle source may be supported by the above discussed compositions of São Jorge olivines (and olivines from Central AZ, in general) and with U-series data (Bourdon et al., 2005; Prytulak & Elliott, 2009).

In detail, the best fit between calculated and observed São João and Manadas basalt compositions is obtained for the 1420 °C melting curve (slightly hydrous source composition). The possibly slightly shallower melting depth of some São João basalts suggested by their lower Gd/Yb (Fig. 12a and b) should be considered cautiously, as these basalts are fairly low in MgO (<6 wt.%) and their trace element budget may have been affected by fractional crystallization processes. Instead, Rosais basalts from the João Dias section tend to be slightly enriched in La/Yb, Gd/Yb and La/Gd, apparently pointing to a melt extraction at slightly higher depth (by ca. 0.1 GPa) and lower melting degree (ca. 1%–2%) compared to the other magmas assuming the mantle source was the same.

However, a contribution of a mantle peridotite with trace element contents slightly different from the primitive mantle may be suggested for Rosais and Manadas basalts. The trace element ratios of Rosais-João Dias basalts (e.g. relatively high Ba/Nb and low Nb/La) may indicate that their mantle source was slightly enriched by small amounts of recycled continental crustal material. The dashed black line in Fig. 12 shows the compositions of melts produced by a peridotite enriched by 0.4% continental crust and overlaps the most enriched Rosais samples. Melt extraction from the enriched source would occur at ca. 2.8 GPa for Rosais-João Dias basalts, i.e. at a similar depth and melting degree to the other V.C. By contrast, the high Nb/La of Manadas and Rosais-Upper Vimes basalts would suggest a source enriched in Nb compared to REE and LILE (the thin black curve in Fig. 12 shows the compositions of melts from a peridotite with 0.9 ppm Nb; Electronic Appendix, Table S4).

The calculated melt segregation depths (ca. 75–85 km, ca. 2.7–2.8 GPa) for São Jorge basalts are higher than the lithosphere-asthenosphere boundary (LAB) in the studied area (which occurs at ca. 1.5 GPa and 50 km under São Jorge; Spieker et al., 2018; cf. Electronic Appendix, Fig. S5). Considering that melting in the asthenosphere generally proceeds until the conductive thermal boundary layer is encountered (Brown & Lesher, 2016), we hypothesize that steep gradients of the LAB depth under the Central AZ (Spieker et al., 2018) may have drained the magmas from the relatively deep melting region to areas of shallower LAB under volcanically active islands (Béguelin et al., 2017). However, U-series data (Bourdon et al., 2005; Prytulak & Elliott, 2009), as well as geochemical differences between, for example, the Holocene lavas from São Jorge (Manadas) and Pico (see below) at distances of less than 30 km (Fig. 1b) argue against large-scale horizontal flows of the melts.

We note that other basalts from the Central AZ have similar REE contents and ratios as São Jorge rocks, suggesting a similar melting regime. On the contrary, basalts from the Western (Corvo and Flores) and Eastern (São Miguel and Santa Maria) AZ tend to have higher Gd/Yb compared to those from the Central Islands and would thus apparently require deeper melting depths. In the case of the Eastern Islands this could be consistent with a significantly thicker lithosphere (cf., Béguelin et al., 2017), but this hypothesis should be considered cautiously as magmas from the Eastern AZ islands issue from a mantle source significantly enriched in incompatible elements (Beier et al., 2018).

Mantle sources

In this section we explore the possible mantle source components contributing to the heterogeneous geochemical compositions of mafic magmas from São Jorge and the Central AZ. Combined trace element and isotopic compositions obtained in this study indicate that São Jorge magmas issued from a significantly heterogeneous mantle source, which is emphasized, for example, by the large range observed in ²⁰⁶Pb/²⁰⁴Pb (18.65 to 20.21), ¹⁴³Nd/¹⁴⁴Nd (0.5128–0.5130) and Δ8/4 values (−64.40 to +45.27; see Table 4; Figs 10 and 11). On the other hand, according to the above-described modelling, the REE variations paradoxically indicate similar melting conditions (i.e. melting depths and degrees) for all São Jorge basalts. Melting of distinct mantle heterogeneities is thus not associated with significant differences in terms of melt extraction depths or temperatures (cf. Stracke & Bourdon, 2009).

According to several studies, a common plume component contributed to or dominated the Sr-Nd-Pb-Hf isotopic composition of the magmas from all AZ islands (Beier et al., 2008, 2018; Béguelin et al., 2017). From this common plume component (circled field, AZ, in Figs 10 and 13), isotopic compositions of the basaltic magmas from single islands spread out to various end-member compositions (Beier et al., 2006, 2007; Elliott et al., 2007; Millet et al., 2009). In Fig. 13, we present a quantitative model to test if the isotopic systematics displayed by São Jorge rocks can be reproduced by a ternary mixture involving three end-members (Fig. 13c and d; end-member compositions in Electronic Appendix, Table S5): 1) a HIMU-type end-member, which is equivalent to the most highly radiogenic ²⁰⁶Pb/²⁰⁴Pb compositions of São Jorge magmas and equivalent to the Terceira-São Jorge end-member of Béguelin et al. (2017); see also Beier et al., 2008); 2) an EM-type end-member, which falls close to the EM-II end-member (as defined for example in Hofmann, 2007, and Weis et al., 2023; and references therein); and 3) the common AZ (CAZ) end-member, which falls within the field of the CAZ plume component reported by Béguelin et al. (2017). Most of the isotopic variability displayed by basalts from São Jorge and from the Central AZ falls within a mixing field defined by these three end-members (Fig. 13; see also Electronic Appendix, Fig. S6).

Manadas

The modelling shown in Fig. 13 suggests that Holocene Manadas magmas derived from a mantle with ca. 90%–40% HIMU-type and 10%–60% of the CAZ component. It should be noted that, when compared with canonical HIMU OIBs (Chauvel et al., 1992; Weis et al., 2023), Manadas basalts have lower ²⁰⁶Pb/²⁰⁴Pb and significantly lower Δ7/4 and higher ⁸⁷Sr/⁸⁶Sr (and Hf and ³He/⁴He isotopic ratios; Béguelin et al., 2017; Moreira et al., 2018).

The Manadas V.C. basalts are strongly enriched in Nb and Ta relative to LREE and they generally have the highest Nb/La and U/La ratios among São Jorge samples. According to the trace element modelling, these magmas formed at ca. 2.8–3.0 GPa from a peridotitic mantle, possibly enriched in HFSE (High Fields Strength Elements) compared to LREE and LILE. Similar isotopic and trace element compositions (e.g. high Nb/La) also characterize all samples from the Vimes section including some from the São João (Lower Vimes, AJ65, AJ66; ca. 1.3 Ma), the Serra do Topo (Middle Vimes, AJ70, ca. 0.7 Ma), and the Rosais V.C. (Upper Vimes, AJ73) and lava flows from the easternmost ridge of the island (near Topo village; data from Millet et al., 2009). Similar compositions are shown by some samples from the other Central AZ islands (Terceira, Pico, Graciosa; Fig. 13; see also the Electronic Appendix, Fig. S7), but not from Faial (²⁰⁶Pb/²⁰⁴Pb < 19.72), and from the Eastern (São Miguel, Santa Maria) and Western Islands (Corvo, Flores; Beier et al., 2010; Madureira et al., 2011; Hildenbrand et al., 2014; Genske et al., 2012, 2016; Béguelin et al., 2017). Recently, Béguelin et al. (2017) modelled the mantle source of AZ lavas with compositions similar to Manadas (their Terceira-São Jorge, TSJ, component overlaps with our HIMU-type end-member). They suggested a contribution from altered recycled and relatively young (Paleozoic) oceanic crust, which also explains the observed isotopic (in particular the relatively low ²⁰⁷Pb/²⁰⁴Pb at quite strongly radiogenic ²⁰⁶Pb/²⁰⁴Pb, i.e. the low Δ7/4) and trace element compositions of the here-studied São Jorge samples.

Rosais

Rosais basalts (0.4–0.1 Ma) from the northern cliffs of the western half of the island (Ouvidor and João Dias section; Hildenbrand et al., 2014; present study) have ⁸⁷Sr/⁸⁶Sr up to 0.7042, ¹⁴³Nd/¹⁴⁴Nd down to 0.51283, low ²⁰⁶Pb/²⁰⁴Pb but high ²⁰⁷Pb/²⁰⁴Pb (Δ7/4 = +2 to +10) and ²⁰⁸Pb/²⁰⁴Pb (Δ8/4 = +2 to + 56) relatively to the NHRL (Figs 10 and 11). Their isotopic compositions could be obtained by mixing 80%–90% of the CAZ component with ca. 10%–20% of the EM-type end-member. Consistently, their trace element ratios (e.g. high Ba/Nb, relatively high La/Gd, La/Yb and the lowest Nb/La of all the volcano-stratigraphic units; Figs 8, 11 and 12) can be modelled by considering melting of a mantle source with very small amounts of recycled continental crust material (<0.5%) added to the peridotitic mantle source. A contribution from the HIMU-type end-member may be possible for some moderately enriched Rosais samples from João Dias (e.g. samples with ²⁰⁶Pb/²⁰⁴Pb ca. 19.5; Figs 10 and 13). However, unlike the Rosais V.C. samples from the north-western cliffs (João Dias and Ouvidor sections), those from the southern and eastern part of São Jorge (including the Rosais-Upper Vimes lavas; this study, Millet et al., 2009; Hildenbrand et al., 2014) have isotopic and trace element ratios rather similar to those from the Manadas V.C. and do not require contribution from the enriched-mantle component. Therefore, the late Pleistocene Rosais V.C. volcanism in general requires contribution from the CAZ plume component mixed with the enriched-mantle end-member for the north-western lavas from São Jorge and with the HIMU-type component for the eastern São Jorge lavas.

Basalts coeval to Rosais rocks from the northern part of the neighbouring Faial Island as well as a gabbro xenolith from Graciosa (Hildenbrand et al., 2014; Larrea et al., 2014) also yield an enriched-mantle flavour, overlapping with the isotopic compositions of western São Jorge Rosais lavas (Fig. 13). In general, Graciosa (1 Ma to Holocene; Larrea et al., 2014; Béguelin et al., 2017) and most Faial basalts (Hildenbrand et al., 2014) plot between compositions of Manadas and Rosais from this study (Fig. 13; see also Electronic Appendix, Fig. S7). On the contrary, there is no evidence of a contribution of EM components in lavas from Terceira and Pico.

EM-II is generally interpreted as originating from deep recycled continental terrigenous sediments (Jackson et al., 2007; Willbold & Stracke, 2010) or from shallow recycling of metasomatized lithosphere (Workman et al., 2004) and is mostly recognized in OIBs from the southern hemisphere (Jackson & Macdonald, 2022). A significant contribution of an EM component is not frequent in central-northern Atlantic OIBs, but enriched isotopic compositions similar to those of Rosais were described for some present-day Central Atlantic MORBs, in particular those sampled next to the Oceanographer Fracture Zone (Dosso et al., 1999) and for Mesozoic and early Cenozoic oceanic basalts (e.g. Mesozoic MORB, New England Seamounts and oceanic sills off-shore Newfoundland from ODP leg 1276; Janney & Castillo, 2001; Hart & Blusztajn, 2006; Merle et al., 2019), Mesozoic continental alkaline rocks from Portugal (Grange et al., 2010) and tholeiitic basalts erupted prior to the break-up of Pangea on the circum-Atlantic continents (Central Atlantic Magmatic Province, CAMP; Callegaro et al., 2014, Merle et al., 2014; Fig. 13). For the latter rocks, the enriched component has been interpreted as resulting from recycling of continental sediments subducted during the Palaeozoic (Merle et al., 2014). The presence of present-day to Triassic basaltic rocks with an enriched-mantle signature in the central-northern Atlantic and adjacent continental margins indicates that this fingerprint persisted in this region since the break-up of Pangea. This signature may be either due to a plume component common in Mesozoic to present-day magmas or to delamination and recycling of continental lithospheric components with such signature acquired during emplacement of the CAMP. However, the available data do not allow to conclusively discriminate between the two options.

São João

A third specific end-member in the São Jorge source is evident in most basalts from the São João V.C. (ca. 1.4–1.2 Ma). These rocks present high ¹⁴³Nd/¹⁴⁴Nd at moderately low ⁸⁷Sr/⁸⁶Sr and ²⁰⁶Pb/²⁰⁴Pb and fairly low ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb (negative Δ7/4 and Δ8/4). The isotopic variations of such rocks can be explained by a mantle source dominated by the CAZ end-member mixed with up to ca. 20% of the HIMU-type component (Fig. 13). A minor contribution of a mantle source similar to that of Manadas seems consistent with the trace element modelling, since São João basalts are slightly enriched in Nb compared to calculated melts derived from a Primitive Mantle peridotite (Fig. 12).

We note that isotopic compositions similar to the CAZ end-member are occasionally found in relatively old Faial (ca. 0.85 Ma; Hildenbrand et al., 2014) and Terceira basalts (ca. 0.4 Ma; Hildenbrand et al., 2014) and in Pico lavas (<0.2 Ma; Béguelin et al., 2017). A few rocks from São Miguel (Sete Cidades Volcano, Beier et al., 2006, 2007) have isotopic compositions, which fall within the field of the CAZ plume field of Béguelin et al. (2017).

Pb isotopic compositions, which plot below the NHRL, at negative Δ7/4, are known for some seamounts and islands from the Central and Northern Atlantic (Tore Madeira Rise, New England, J Anomaly Ridge, Vesteris seamount, Canary, Madeira and Iceland islands, Fig. 13; Mata et al., 1998; Geldmacher & Hoernle, 2000; Geldmacher et al., 2006, 2008; Merle et al., 2019; Taylor et al., 2020; Harðardóttir et al., 2022; Beloša et al., 2024) and in MORBs sampled at the Oceanographer Fracture Zone (Dosso et al., 1999; Fig. 14). However, most of these oceanic basalts (except those from the J-Anomaly Ridge; Merle et al., 2019) are characterized by positive Δ8/4 and by Sr-Nd isotopic compositions different from those of São João. According to Geldmacher et al. (2008), the EM-I-like isotopic composition of Godzilla Seamount samples reflects recycling of metasomatized Archean continental lithospheric mantle in their source (cf. Salters & Sachi-Kocher, 2010). Merle et al. (2019) suggest a shallow origin to explain the common isotopic fingerprint of Mesozoic Central Atlantic seamounts. However, the presence of this component in the São João V.C. and in several Central AZ islands (Béguelin et al., 2017; Beier et al., 2018 and references therein) indicates that this is a component of the AZ plume and may thus be of deep origin.

Space- and time-related evolution of Central AZ magmas from a heterogeneous plume

Our geochemical and geochronological data for São Jorge allow us to constrain the spatial and temporal distribution of source heterogeneities. Basaltic magmas dominated by the CAZ and the EM-type components, respectively, were erupted between ca. 1.37 and 1.21 Ma in the eastern portion of São Jorge (São João and lowermost Vimes sections) and between 0.41 and 0.22 Ma in the north-western portion of the Island (João Dias and Ouvidor sections). In both cases, eruption of magmas with ‘anomalous’ composition persisted for ca. 0.2 Ma and was then replaced by eruption of magmas dominated by the HIMU-type component. As previously discussed, magmas similar in composition to São João basalts were diachronously erupted during probably initial volcanic phases of other Central AZ Islands, while eruption of Rosais-like magmas occurred synchronously in Faial and possibly on Graciosa (see the map in Fig. 14a for the distribution of these magmas in the Central AZ). Our interpretation is that the HIMU-type component was the dominant plume component in the Central AZ at least since 1.37 Ma, while the CAZ and EM-type end-members were only locally present and were exhausted within ca. 0.2 Ma (Fig. 14). These anomalous compositions seem to be derived by partial melting of a heterogeneity within the AZ plume characterised by a quite restricted vertical length estimated at ca. 5–10 km considering the postulated upflow velocity of the AZ plume (ca. 3–4 cm/year; Bourdon et al., 2005; note that this calculated length is a minimum estimate as we did not consider plate motion and horizontal flow of the plume mantle towards the Mid Atlantic ridge; O’Neill & Sigloch, 2018). The horizontal shapes of these anomalies are difficult to constrain, considering that available geochemical and geochronological data are mostly limited to the sub-aerial volcanoes and that there are no clear constraints on the extent of the melting zone. Nonetheless, both the magmas with EM-type signature and with compositions similar to São João seem to have been erupted in relatively thin NNE–SSW-directed, elongated belts (Fig. 14b). This spatial organization points towards a filament-shaped distribution of the heterogeneities within the plume. Notably, the apparent NNE–SSW elongation of these filaments is parallel to the main direction of the horizontal shear within the shallow mantle (Adam et al., 2013), suggesting that the filaments may have been stretched by mantle flow. Filament-shaped geochemical anomalies with limited vertical extent are known from other OIBs (Cordier et al., 2021) and could result from a small viscosity contrast between the filaments and the surrounding plume material (Farnetani et al., 2018). Such a weak viscosity difference could be yielded if the major element composition of both the dominant plume material and the filament anomalies are similar, with a lithology dominated by a peridotite (and not pyroxenitic) matrix. This appears to be the case for the Azorean plume, as pointed out by the low-Ni olivine compositions of São Jorge and other AZ Islands, by the mantle-melting modelling presented in this study (Fig. 12), and by U-series data (Bourdon et al., 2005; Prytulak & Elliott, 2009).

Finally, we note that according to this and several previously published studies (Béguelin et al., 2017; Beier et al., 2018 and references therein) the Central AZ OIBs seem to be at least partially different from those from the Eastern AZ. In particular, the dominant mantle component in the Central AZ appears to be HIMU-type, a component which seems to be absent in the Eastern AZ (Fig. 13). Moreover, the EM-type component, which apparently contributed to the late Pleistocene Rosais and coeval magmas from Faial and Graciosa was detected only in the Central AZ. On the contrary, as highlighted by Béguelin et al. (2017) and Beier et al. (2018), basalts from the Eastern, Central and Western AZ display a common end-member composition, which we have identified as being dominant only for the early volcanic phase at São Jorge (São João V.C.). According to a tomographic study of the shallow mantle in the AZ region (Yang et al., 2006), the hot rising plume under the AZ seems to be divided into two branches at ca. 200 km depth (Fig. 14c), one branch rising towards the Central AZ and the other one towards São Miguel in the Eastern AZ. These two branches seem to produce magmas with partially distinct isotopic compositions as previously described.

On a larger scale, an aspect that emerges from the present and previous studies on the AZ is that there is some overlap in Sr-Nd-Pb isotopic compositions with OIBs from the Central Atlantic (e.g. the Canary and Madeira archipelagos, Fig. 13a and b). The melting degrees and melting depths calculated in this and previous studies (Beier et al., 2018) are similar to those calculated for the Canary and Madeira mafic magmas (Taylor et al., 2020). This invariance of melting parameters is consistent with the small difference in terms of plume buoyancies observed between the AZ and Canary archipelagos (0.38 and 0.29 Mg s⁻¹, respectively; King & Adam, 2014). Through time, multiple mantle-upwellings from a common plume source might have triggered several episodes of volcanism at Madeira, Canary and, possibly, the AZ. The common plume source may be related to the seismically slow ‘Central-East Atlantic Anomaly’ (Saki et al., 2015; Civiero et al., 2021) stacked below the 660 km discontinuity and being the north-western branch of the mantle upwelling rooted at the Africa (or Tuzo; Torsvik et al., 2014) ‘large low-shear velocity province’.

However, if melting conditions appear to be broadly invariant between these three Macaronesian archipelagos, the extent of Sr-Nd-Pb isotopic heterogeneity of their mantle source differs. As shown in Fig. 13, the extreme variability observed for the Azorean basalts (this study, Béguelin et al., 2017; Beier et al., 2018 and references therein) contrasts with the relatively narrow isotopic spread of Madeira and Canary OIBs (Mata et al., 1998; Geldmacher & Hoernle, 2000; Taylor et al., 2020 and references therein). Strongly variable isotopic compositions also characterize MORBs at latitudes 35–40° North, in particular those sampled at the Oceanographer Fracture Zone (Dosso et al., 1999), and Central Atlantic seamounts generally located next to this fault (Geldmacher et al., 2008; Merle et al., 2019 and references therein; Figs 1a and13) pointing to a more heterogeneous composition of the plume or to entrainment of shallow mantle components in this area.

CONCLUSIONS

In this work we show that the subaerial part of the Island of São Jorge was built during four successive phases separated by variable periods of volcanic quiescence. These phases are represented by the São João, Serra do Topo, Rosais and Manadas V.C.s (Figs 1, 3 and 14). Volcanism is exclusively fissural and controlled by tectonics, with the alignments of cones coinciding with major faults. New ⁴⁰Ar/³⁹Ar ages, a geomagnetic survey, geomorphological and fieldwork observations confirm the presence of the older São João volcano-stratigraphic unit, previously reported by Hildenbrand et al. (2008), as well as the presence of the Serra do Topo V.C. (ca. 0.8–0.5 Ma) and Rosais V.C. in the eastern sector of São Jorge, originally attributed to a single stratigraphic unit (Topo V.C.). Rosais V.C. flows (ca. 0.4–0.1 Ma) are widespread in the western and are also present in the eastern portion of the island. Radiocarbon dates indicate a Holocene age (<8 ka) for the Manadas V.C.

In general, São Jorge magmatism seems to be derived from a mantle source, which is locally heterogeneous and displayed significant changes over time. The oldest rocks (São João V.C.) show Pb isotopic compositions at values lower than the NHRL and similar to the CAZ plume component proposed by previous studies. The late Pleistocene Rosais basalts from the João Dias section display an enriched geochemical mantle signature, while the Holocene magmatic rocks (Manadas V.C.), along with some Pleistocene lavas from the eastern half of the island show a HIMU-type fingerprint. The eruption of basalts dominated by either the São João or the EM-type Rosais compositions was spatially and temporarily limited. Magmatic rocks with compositions similar to those of the São João and the enriched Rosais lavas are occasionally present in some coeval lavas from neighbouring islands (Central AZ), suggesting that narrow and short-lived filaments of heterogeneous composition were present in the Central AZ mantle plume. The dominant composition within this plume was probably similar to that of the Manadas lavas and is widespread on São Jorge since 1.37 Ma and on the other Central AZ islands, while it is not known for basalts from the Eastern AZ, which were probably produced by a distinct branch of the AZ plume.

Supplementary Data

Supplementary data are available at Journal of Petrology online.

Acknowledgements

The present work is a contribution to projects SHA-AZ (PTDC/CTE-GIX/108637/2008) and UIDB/50019/2020 funded by Fundação para a Ciência e a Tecnologia, I.P./MCTES through national funds (PIDDAC) and by GRICES (Portugal)—CNR (Italy) Bilateral Cooperation project ‘A província magmática do Atlântico Central (CAMP) em Portugal’ and by GRICES (Portugal)—CNRST (Morocco) Bilateral Cooperation project ‘Magmatismo intraplaca em domínio continental e oceânico’. S.C. acknowledges funding from the Research Council of Norway (Young Talent grant 301096), A.M. and A.D. acknowledge support from Ministero dell’Università e Ricerca (PRIN, 20178LPCP). J.M., J.M., A.B.S. and P.S. acknowledge the financing by the Portuguese Fundação para a Ciência e a Tecnologia (FCT) I.P./MCTES through national funds (PIDDAC)—UIDB/50019/2020, UIDP/50019/2020 and LA/P/0068/2020. Detailed and constructive reviews by R. Merle, P. Béguelin, J. Geldmacher, H. Kawabata, and an anonymous reviewer and editorial handling by S. Nielsen provided very helpful comments on the original manuscript.

Data availability

All new analytical data of whole rock and in situ mineral analysis are provided in Supplementary Appendix Tables S1 to S3 and deposited at EarthChem repository with following DOI: https://doi.org/10.60520/IEDA/113519 and https://doi.org/10.60520/IEDA/113499.

References