Spatial controls on the generation of low-δ18O basalts on São Miguel, Azores

Volcanism in the Azores is generally accepted to result from the decompression melting in the upper mantle associated with the Azorean plume. Basalts on the island of São Miguel have been previously recognised as having δ 18 O values lower than MORB with different authors ascribing this low-δ 18 O signature to either a component within the plume itself or as the result of assimilation of crustal materials at shallow level. Here we assess these contrasting models by using a suite of samples that are both geographically and temporally spread. We find that while low-δ 18 O mafic magmas (reaching 4.98 ‰) do occur on São Miguel, they are spatially restricted to the Picos fissure system, between the central volcanoes of Sete Cidades and Fogo. Basalts of similar age from outside this restricted region return δ 18 O values that are similar to MORB. This spatial constraint argues that the source of the observed low-δ 18 O signature is within the crust rather than a component of the plume. Previous studies on mafic to intermediate rocks of the Picos fissure system have identified the assimilation of broadly syenitic lithologies, which were found as co-erupted lithic clasts at the Fogo volcano and may have δ 18 O values as low as 2.6 ‰. However, trace element geochemistry indicates that these lithologies have not been assimilated within our low-δ 18 O samples, a more likely assimilant is altered oceanic crust or mafic lavas within the island edifice.


Introduction
Mantle plume magmatism has occurred in a punctuated fashion throughout Earth's history.Such events have been fundamental in shaping the evolution of our planet in many ways, including controlling continental rifting, global climate, and even mass extinctions (Koppers et al. 2021).The volcanic provinces resulting from plume magmatism may persist for tens of millions of years, typically with brief periods of high-volume extrusion (e.g.Storey et al. 1995) but considerable variability in their expressions exists related to the properties of the mantle upwelling and the crustal rocks upon which it impinges.Gaining clarity on the physical and compositional properties of these upwellings or plumes remains a challenge.Geophysical studies have made significant progress in imaging thermal anomalies within the mantle, with a range of structures J o u r n a l P r e -p r o o f Journal Pre-proof observed persisting to various depths in the mantle (many studies well summarised by Koppers et al. 2021).Complementing these findings, a range of geochemical studies have addressed the mineralogical makeup of mantle plumes and their sources (e.g.Hauri 1996;Sobolev et al. 2005;Barker et al. 2023).Isotopic systems have also been employed to address questions such as inputs to the plume itself (e.g.Beguelin et al. 2017), temporal changes in the magma source (Cousens et al. 1993;Abouchami et al. 2005) or provinciality within a single upwelling (Shorttle et al. 2013).
With such a range of potential behaviours and settings, investigating endmember types has clear benefit, particularly when magmatic systems are still active.The Azores mantle upwelling represents one such end-member with a high inferred water content, relatively muted excess temperatures, and low 3 He values (Beier et al. 2012;Moreira et al. 2012;Métrich et al. 2014).In addition, magmas in the Azores exhibit large ranges in radiogenic isotopic compositions for ocean island settings, often with coherent variations across the archipelago (e.g.Moreira et al. 1999;Elliott et al. 2007, Millet et al. 2009).Many of these isotopic studies have identified an ancient component (estimated as >2.5 Gyr) as a notable constituent within the Azores plume (Schaefer et al. 2002;Beier et al. 2007;Elliott et al. 2007;Béguelin et al. 2017).
Oxygen isotope compositions of the Azores magmas have also been used to provide constraints on the componentry of the plume.Some studies, focussing on O isotopes, have suggested that the Azores plume itself is low-δ 18 O (Widom and Farquhar 2003;Turner et al. 2007) while others have suggested that the low-δ 18 O signature is inherited in the shallow crust (Genske et al. 2013).Similar differing interpretations exist for other light stable isotopes with Turner et al. (2007) attributing variable B isotopes to a plume source, while Genske et al. (2014) suggest variations in B and Li isotopes originate from shallow assimilation.
On a smaller scale, it has been noted that there may be considerable variability on a single island, particularly the case for São Miguel, which is the focus of this study (Widom et al. 1997;Haase and Beier 2003;Elliott et al. 2007;).Haase and Beier (2003) found distinct Sr isotopic compositions from basalts erupted in 'the waist zone' (also known as the Picos fissure system) between Sete Cidades and Água de Pau (Fogo) volcanoes (Fig. 1).They attributed these variations to small-scale heterogeneities being sampled due to the tectonic structure of the lithosphere underneath the island.Similarly, using Fe isotopes, Ruttor et al. (2022) concluded that the heavier Fe isotopic compositions found in western São Miguel were the result of addition of a MORB-like component to an initial OIB-like basalt due to overprinting within the active Terceira Rift.
Despite the known spatial isotopic variability in São Miguel samples, the vast majority of samples previously studied for δ 18 O that could be reliably located were taken from the centre-west of the island, the Picos fissure system (Fig. 1).With this work we expand the existing dataset from São Miguel both in terms of geographic and temporal spread of samples, and by the analysis of multiple mineral phases per sample.With this approach, we aim to investigate whether deeper or shallower processes control δ 18 O values on São Miguel.

Geological Background
The Azores archipelago (North Atlantic Ocean) consists of nine islands (Fig. 1 inset) straddling the Mid-Atlantic Ridge.Its complex geodynamic setting results from the interaction between the triple junction of the North American, Eurasian, and Nubian lithospheric plates and a mantle melting anomaly, often referred to as the Azores mantle plume (e.g.Cannat et al. 1999;Storch et al. 2020).All the islands are of volcanic origin and seven of them have active volcanic systems, most of which have erupted in historical times, that is within the last 500 years.Overall, volcanic rocks of the Azores islands show a full range of compositions from basanite to trachyte and more rarely rhyolite (Larrea et al. 2018).
The island of São Miguel, the largest and most populated of the Azores is the site of this study.Some 60 km long and 12 km wide, São Miguel hosts seven volcanic systems: three active central volcanoes with calderas (Sete Cidades, Fogo, -also known as Água de Pau, and Furnas) linked by two active fissure systems (Picos and Congro), and two extinct volcanic systems (Povoação and Nordeste) in the eastern part of the island (Gaspar et al. 2015) (Fig. 1).This study focuses on the most prominent regions of basaltic volcanism, which are briefly described below.
Sete Cidades volcano, the westernmost of the three active central volcanoes of São Miguel, is a low-relief edifice (845 m a.s.l.) characterised by a 5 km wide sub-circular caldera with steep-sided walls 30-400 m high.Its lower volcanic pile dates back to > 210 ka (Moore 1990) and is predominantly formed by basaltic and a few trachytic lavas alternating with pyroclastic deposits.The upper sequence (< 36 ka) includes three major trachytic pyroclastic deposits at 36 ka, 29 ka, and 16 ka (Queiroz et al. 2015) related to different phases of formation of the caldera, which are intercalated with subordinate trachytic and basaltic products.Rocks of intermediate compositions are less common.The recent, < 5 ka, products comprise trachytic pumice deposits from 17 intracaldera eruptions (Pepom tephras, Fig. 2A) and basaltic lavas and scoriae from 12 flank eruptions (Queiroz et al. 2015).
Picos fissure system also known as 'the waist zone' is, at 7.5 to 12 km wide, the narrowest part of the island linking Sete Cidades in the west to Fogo in the east.This low-lying area (up to 485 m a.s.l. in the central part) is characterised by two sub-parallel volcanic ridges of scoria cones, whose lavas gently slope to the north and south coasts.The Picos fissure system is the youngest volcanic area of São Miguel, the age of the oldest products is poorly constrained to >30 ka (Moore 1990).The older volcanism was widespread across the entire fissure system, while the recent activity (< 5 ka) has been concentrated in the two main ridges, where at least 30 predominantly basaltic eruptions have occurred (Ferreira et al. 2015) with subordinate volumes of hybrid magmas (Storey et al. 1989).
Nordeste volcanic system, reaching 1162 m a.s.l., corresponds to a highly eroded large shield volcano that constitutes the eastern end of São Miguel Island.It is the oldest (> 878 ka;Johnson et al. 1998) volcanic edifice on the island and is predominantly formed by a sub-parallel stack of basaltic lavas.The older stratigraphy of this extinct shield volcano is poorly known, but the upper volcanic pile, 820 to 780 ka, is composed of ankaramitic and basaltic lavas (Fig. 2B), with a small volume of trachytic lavas found atop.Rocks of intermediate compositions are rare (Duncan et al. 2015).

J o u r n a l P r e -p r o o f
Journal Pre-proof

Methodology
Fresh mafic volcanic products were sampled from across the island of São Miguel (Fig. 1) in an attempt to cover the spatial and temporal variability of mafic volcanic products.After removing any unwanted surficial material, samples were crushed with a vertical milled, then sieved for mineral separation and another aliquot being powdered for bulk rock analysis using agate mortars.Samples were prepared for X-ray fluorescence (XRF) at ETH Zürich following standard methods as described in Cortes-Calderon et al. (2022).Following XRF analyses, glass pills were broken and internal surfaces were analysed via laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS), also at ETH Zürich.Typical relative errors are <1% for the majority of major elements and <2% for Na2O.For trace elemental analyses by laser ablation, uncertainties are typically < 1ppm for elements < 10 ppm and <20 ppm for elements < 1000 ppm based on multiple analyses of the BCR-2 reference material (Jochum et al. 2005).All new geochemical data are provided in the supplementary materials.
Where present olivine crystals were separated from the bulk samples to investigate crystal textures and compositions, particularly in light of earlier studies (e.g.Genske et al. 2013).Crystals were picked under a binocular microscope, mounted in epoxy and polished to 1 micron prior to carbon coating.Backscattered electron images using a JEOL JSM-6390 LA Secondary Electron Microscope (SEM) were obtained at ETH Zürich (a library of olivine images is provided in the supplementary data).Following zonation patterns as observed in backscatter electron imaging, olivine crystals were then analysed using a JEOL-JXA 8230 electron probe microanalyser (EPMA) equipped with five tuneable wavelength dispersive spectrometers hosted at ETH Zürich.Measurements were performed following methodology described in Ellis et al. (2022) at 15 kV acceleration voltage, 20 nA beam current and focused beam.The standards employed were rutile for Ti, pyrolusite for Mn, chromite for Cr, bunsenite for Ni, albite for Na and Si, forsterite for Mg, anorthite for Al and Ca, fayalite for Fe, and microcline for K. On-peak acquisition time was set to 30 s for all elements except Fe, Mn and Ni that were acquired for 20 s. Background was modelled using the mean atomic number (MAN) background fit correction for each element using standards that do not contain the analyte (Donovan and Tingle 1996).Data was acquired and processed using probe for EPMA software (Donovan et al. 2021).
Other crystals, depending on the sample mineralogy, were also picked under a binocular microscope, typically from the 355 to 710 micron size fraction to be analysed for O isotopes by laser fluorination at the Department of Geological Sciences at the University of Cape Town, South Africa.Analyses typically involved 2-4 mg of material (2-5 crystals) and followed the methodology of Harris and Vogeli (2010) as modified in Ellis et al. (2022).The oxygen produced was converted to CO2, the amount of gas was measured by reference to the mass 32 peak voltage to check for complete conversion and the

J o u r n a l P r e -p r o o f
Journal Pre-proof isotopic ratios were measured using a DeltaXP gas-source mass spectrometer in dual-inlet mode at the University of Cape Town, South Africa.
We strived wherever possible to analyse two mineral phases per sample to allow testing for high-temperature equilibrium using well-established fractionation factors.In a number of cases we analysed the same samples in duplicate during different analytical sessions.Over the course of the three years of analyses involved in this study the long-term average difference between duplicates of the Monastery garnet (MON GT) standard (δ 18 O = 5.38‰, Harris and Vogeli 2010) was 0.12‰ (n = 393 pairs), corresponding to a 2σ value of 0.16‰.

Bulk rock geochemistry
Mafic samples from São Miguel generally have higher and more variable crystal contents than the evolved lithologies.Some of the Nordeste mafic lavas have large and abundant olivine and clinopyroxene crystals occurring both as single grains and as nodules of cumulates (Fig. 2B).
All of the samples considered in this study are mafic with recalculated anhydrous SiO2 between 43.7-53.7 wt.%, and total alkalis between 2.08-8.32wt.%.These values are consistent with the compositions that volumetrically dominate the island of São Miguel (Moore 1990).Most notable are the samples from Nordeste that contain conspicuous accumulations of olivine and clinopyroxene (Fig. 3) that are observable in their lower total alkalis, elevated MgO contents (to nearly 16 wt.%)and the high abundances of Ni and Sc, while simultaneously being depleted in elements that are incompatible in clinopyroxene and olivine such as Ba and Zr (Fig. 3).Following Beier et al. (2006) we use a cutoff point of 12 wt.%MgO to distinguish between mafic samples without cumulate olivine addition and those that are accumulative in olivine and clinopyroxene, with Ubide et al. (2022) also suggesting a similar approach using a cutoff at 10 wt.% MgO.The three compositions from Nordeste that are significantly above the Beier et al. (2006) cutoff (13.31, 13.78, and 15.97 wt.% MgO) are those that show hand specimen evidence for the mechanical entrainment of mafic cumulates of olivine and clinopyroxene (Fig. 2B).Some lavas in the Picos fissure system (Fig. 2D) are also relatively high in MgO (9.32 wt.%) and contain a prominent crystal cargo.Trace element ratios for the samples in this study are broadly similar to those described by Beier et al. (2006) with Ba/Rb higher for Sete Cidades and Picos samples than for Nordeste and the reverse true for Rb/Sr.Although we note that there is more variability in our samples than those of Beier et al. (2006).Chondrite-normalised REE plots (McDonough and Sun 1995) have Eu/Eu* (Eu/Eu*= EuCN/(SmCN*GdCN) 0.5 ) ranging between 1.02 and 1.14, consistent with the minimal role of plagioclase fractionation in this compositional range (Fig. 3) and a general enrichment in REE.

J o u r n a l P r e -p r o o f
Journal Pre-proof

Olivine textures and compositions
Olivine is common in the mafic samples from the studied volcanic systems of São Miguel.In Nordeste, samples often contain cumulate nodules of olivine and clinopyroxene reaching 10 cm in diameter (Fig. 2B) and with total crystallinity of the samples on the order 15-30%.In addition, the crystals found both as nodules and as single grains may reach prodigious size (1 cm or more).In contrast, some of the mafic samples from the Picos fissure system are almost aphyric.Olivine zonation styles vary within samples and examples are provided in Fig. 4 (and supplementary figure 1).Even within a volcanic system the olivines can vary significantly.From the Sete Cidades volcano, olivines from the Ferrarias scoria (AE AZ 1918) intercalated within the Pepom tephras (Fig. 2A), have a distinctly brighter, sharply defined, 10-20 µm thick rim in BSE imaging while the other samples from Sete Cidades contain texturally homogeneous olivines (Fig. 4D).
São Miguel olivine compositions range between Fo65-88 and agree well with the range of Azores olivine compositions published by Sobolev et al. (2005), albeit with some of the Nordeste compositions more Ni-rich at a given Fo content.Overall, the olivines are normally zoned (Fig. 4A) with the largest intragrain variation of Fo65-86.The olivines from Sete Cidades are notably more homogeneous than those from the Picos or Nordeste systems and even within the Picos and Nordeste individual samples show variable zonation patterns (Figs.4E-G).Few olivine compositions pass a classic equilibrium test (Fig. 4B) on a 'Rhodes diagram' (Roeder and Emslie 1970).Overall, in any given sample the range of olivine compositions suggests a role for mixing between mafic magmas as previously suggested (Beier et al. 2006).The lack of equilibrium in the more mafic samples is likely a combined result of samples containing abundant cumulate crystals (e.g.Fig. 2) and the use of bulk rock compositions for the liquid rather than glass.The three samples from Nordeste that were suggested to be accumulative in olivine are those with the highest Mg-number and the fewest olivine compositions that approach equilibrium (Fig. 4).We note that Widom and Farquhar (2003) found a similar range of olivine compositions overall and a comparable range within a single sample to those we report here.The two samples from Widom and Farquhar (2003) that overlap with the accumulative samples from Nordeste have MgO contents of 11.48 and 11.88 wt.% and may also have 'excess' olivine.

Oxygen isotopic compositions
Measured δ 18 O values range from 4.58 ‰ from an olivine from the Picos fissure system to 6.66 ‰ from a feldspar from a basaltic scoria from Sete Cidades (all new data in supplementary material and supplementary figure 3).To estimate melt δ 18 O values we make the following adjustments to mineral data: +0.4 ‰ for olivine (Genske et al. 2013), -0.2 ‰ for feldspar, and +0.2 ‰ for clinopyroxene (Harris et al. 2000).Using these values return estimated magmatic δ 18 O values of 4.98-6.46‰ (Fig. 5).Analyses of co-existing mineral phases from a single sample indicate that the O isotopic compositions reflect high-temperature equilibrium (Fig. 5) and multiple measurements of the same mineral phase from J o u r n a l P r e -p r o o f Journal Pre-proof the same sample reproduce well (supplementary figure 2), even allowing for the presence of small inclusions, particularly within the mafic phases.Where multiple mineral phases were analysed within a single sample, we take the magmatic value from the crystal phases in the following order olivine, clinopyroxene, feldspar.Where olivine and clinopyroxene have been analysed from the same sample, Δ 18 Ocpx-olivine values range between -0.22 and 0.48 ‰ with the average 0.15 ‰, approximating analytical uncertainty.Across the dataset as a whole (i.e.including samples with analyses from only olivine or clinopyroxene in the average) the overall average magmatic value calculated from olivine is <0.02 ‰ higher than the average value calculated from clinopyroxene.Estimates from the literature for the Δ 18 Ocpx-olivine are similar to our results, with estimates of ~0.40 ‰ (Mattey et al. 1994), between 0.3 and 0.4 ‰ (Chazot et al. 1997), between 0.04 and 0.29 ‰ (Dobosi et al. 1998), and between 0.24 and 0.40‰ (Day et al. 2012).We consider the 6.46‰ value from a scoria deposit feldspar with caution, as the other Sete Cidades samples analysed indicate magmatic values estimated from feldspar are somewhat higher than those from olivine and clinopyroxene (order 0.4-0.6 ‰).

Low-δ 18 O magmas on São Miguel
The O isotopic composition of a volcanic rock is the final aggregated result of all the processes having occurred during its generation and traverse of the crust (Bindeman 2008;Troch et al. 2017;Rooyakers et al. 2023).It is widely accepted that closed-system fractional crystallisation induces a modest increase in δ 18 O from the values of 5.4-5.9‰ generally assumed to reflect 'mantle' values (Eiler 2001;Bindeman 2008;Troch et al. 2020).To define a magma as low-δ 18 O, therefore depends upon establishing deviation from reference δ 18 O trajectories resulting from closed-system behaviour.To do this, we follow the model of Troch et al. (2020) that summarises the various estimates of closed-system behaviour and places a lower limit on what 'normal-δ 18 O' magmas should be (Fig. 5A).Differentiation in Fig. 5 is evaluated by magma SiO2 content.With our data being lavas or bulk juvenile tephra there is always the potential that some compositions may not reliably reflect melts when crystal contents are high.The Nordeste compositions suggest that they contain cumulate olivine and / or clinopyroxene, but such features are predominantly observed in trace element data rather than SiO2 (Fig. 3).
The generation of a low-δ 18 O magma is widely accepted to require the assimilation of a hydrothermally altered lithology that underwent alteration at high temperatures (Bindeman 2008).The δ 18 O imprint on the altered rocks will be further controlled by the δ 18 O of the water involved.The variation in δ 18 O of meteoric water can be significant at different regions across the globe.The meteoric water in the Azores is less negative (-4.1‰ to -5.2‰, Woitischek et al. 2017) than the values reported from areas known for recent low-δ 18 O magmas

J o u r n a l P r e -p r o o f
Journal Pre-proof (e.g.Yellowstone -19 ‰ Hildreth et al. 1984).Thus the potential to generate low-δ 18 O magmas varies across settings.
A low-δ 18 O signature in a magma could have a variety of potential origins, which can be broadly categorised as: i) a primary signal of the mantle plume with the low-δ 18 O values inferred to represent an addition of high-temperature hydrothermally altered slab material to the plume source (Widom and Farquar 2003;Turner et al. 2007).ii) a result of assimilation fractional crystallisation (AFC) processes either progressively during magmatic evolution (Genske et al. 2013), or as part of a two-stage processes where the evolved magmas assimilate low-δ 18 O materials (Hampton 2021).iii) direct melting of hydrothermally altered rocks, either partially or totally, with the resultant melt then erupted (Bindeman and Valley 2000).Where the lowδ 18 O signature is observed in mafic magmas, the direct melting mechanism is problematic due to the requirements of melting an altered lithology while having the resultant melt be mafic.If the altered materials were initially mafic it is unlikely that they would melt as a result of mafic magma recharge (e.g.Ellis et al. 2023), while if alteration had changed the composition of the rock to something more readily 'meltable' then the resultant composition would be unlikely to be basaltic.Experimental petrology studies of hydrothermally altered mafic lithologies indicate that partial melts of such rocks typically produce silicic melts (Beard and Lofgren 1991;France et al. 2010).For these reasons we do not consider this option further.

A plume-related origin
Previous studies have produced conflicting views on the nature of the O isotopes within the Azores magmas.Widom and Farquhar (2003) and Turner et al. (2007) ascribed the low-δ 18 O values they found predominantly from the Picos fissure system to a low-δ 18 O component of the Azores mantle upwelling with the plume source estimated to have >10% of hydrothermally altered oceanic crust (Widom and Farquhar 2003).Genske et al. (2013) however, found by coupling O isotopes and major elemental compositions of the same olivine crystals from samples across the Azores that lowering of the δ 18 O tracked the decreasing Fo content of the olivines, i.e. lowering of the δ 18 O occurred during magmatic evolution.Regressing this trend back to the most primitive compositions gave results similar to the typical 'mantle olivine' value of δ 18 O of 5.2 ‰ (Eiler 2001).
Olivine δ 18 O values from this study range from 4.58 to 5.36 ‰ (n = 18) almost identical to the published olivine values of 4.57 to 5.28 ‰ (Fig. 6).However, when plotted by location, for samples that could be reasonably located, it is notable that the lowest δ 18 O values cluster within the Picos fissure system (Fig. 6A) the central region of the island dominantly sampled by Widom and Farquhar (2003).Our Picos olivine values average 4.85 ± 0.25 ‰, 1 std.dev., n= 8 in good agreement with the overall average of Widom and Farquar (2003) of 4.87 ± 0.15 ‰, 1 std.dev., n= 8.In the east and west sectors of the island however, olivines tend to have δ 18 O values closer to accepted mantle values (5.2 ± 0.2‰, Eiler 2001).Taking all olivine analyses in Table 1, including repeats from a single unit, Sete Cidades olivine averages 5.14 ± 0.12 ‰, 1 std.dev., n= 6 and Nordeste olivine averages 5.17 ± 0.13 ‰, 1 std.dev., n= 5.The variability in

J o u r n a l P r e -p r o o f
Journal Pre-proof the δ 18 O of the mafic magmas reported here may reflect a number of potential processes.A first possibility is that the assimilant required to lower the δ 18 O in the mafic magmas simply does not exist under the central volcanoes.A second option is that the mafic rocks in the fissure systems reflect a relatively earlier stage of the magmatic life of that system.In this scenario, the assimilant is progressively exhausted (similar to the progressive inhibited crustal assimilation model of Meade et al. 2014) and therefore the young mafic magmas erupted from the central volcanoes have a lower potential for assimilation.To distinguish between these hypotheses would require access to the older mafic rocks of the central volcanoes, which are typically poorly exposed if at all due to the frequent resurfacing at these active centres.However, we do note that in the Nordeste system where we have samples throughout the stratigraphy, there is no correlation between δ 18 O value and stratigraphic position.
----INSERT FIGURE 6 olivine map figure ----In contrast to the spatial relationship observed in magma δ 18 O values, there is little to no relationship in terms of age.Shown in Figure 6B are the olivine δ 18 O values plotted as a function of sample age.This includes both our data and those from the literature where samples as located in the original work have been attributed to map units of Moore (1991).Where samples have the same position on the X-axis on this diagram they are thought to be from the same unit, with good agreement with earlier work.While the samples from Nordeste are significantly older (on the order of hundreds of thousands of years), the eruption of low-δ 18 O and normal-δ 18 O magmas from Picos and Sete Cidades contemporaneously indicates the δ 18 O character is unrelated to the age of the sample and is unlikely to reflect a source characteristic.
----INSERT FIGURE 7 18 O vs geochemistry---- Haase and Beier (2003) noted that the Sete Cidades and the Fogo volcanoes were isotopically distinct despite being only some 20 km apart, with Sete Cidades having 87 Sr/ 86 Sr < 0.7038 while Fogo-related lavas have 87 Sr/ 86 Sr > 0.7040.They interpreted these differences as reflecting separate magmatic systems fed by different mantle sources.However, when plotting the Widom and Farquhar (2003) 87 Sr/ 86 Sr data against δ 18 O it is clear that variations in δ 18 O occur independently from the radiogenic isotope composition of a sample (Sup.Fig. 4).On a mineral-scale, Genske et al. (2013) showed a relationship between Fo content of the olivine (as a proxy of magmatic evolution) and δ 18 O from the Azores archipelago, suggesting that assimilation occurred as evolution progressed.Our results are broadly in agreement with this (Sup.Fig. 5) but our lowest δ 18 O values are off the trend presented by Genske et al. (2013).However,

J o u r n a l P r e -p r o o f
Journal Pre-proof given that potential assimilants could vary widely in their δ 18 O values, a similar extent of AFC could result in very different δ 18 O values for a given Fo content.

The role of the crust
At the Sete Cidades caldera system, the youngest trachytes, the Pepom tephras, have magma δ 18 O values mostly between 5.03-5.80‰ with a single value extending down to 4.13 ‰ (Ellis et al. 2022).These values are lower than the young mafic magmas in the same system, some of which are intercalated within the trachytic Pepom tephra stratigraphy (Fig. 2, 8).While models (Bindeman et al. 2004;Harris et al. 2000;Bucholz et al. 2017) differ in the extent of the increase in δ 18 O during closed-system fractionation leading to a silicic magma, all models agree that this process causes an increase in δ 18 O.Thus, assuming that the mafic products intercalated with the silicic deposits at Sete Cidades reflect suitable parental magmas for the trachytes, this drop in δ 18 O occurs during magmatic evolution and likely within the island crust.
----INSERT FIGURE 8 of just Sete Cidades----Similarly, the spatial restriction of mafic, low-δ 18 O values in the Picos fissure system also implies a role for crustal assimilation.The Picos fissure system, or the 'waist zone', was shown by Storey et al. (1989) to contain numerous 'hybrid' mafic lavas with partially resorbed anorthoclase crystals.These hybrid lavas are readily identified in bulk geochemistry by their elevated K/Zr at a given Ti/Zr (Storey et al. 1989;Fig. 9) that results from the assimilation of syenites / alkali feldspar-dominated cumulates.The low-δ 18 O values we identify in some of the Picos basalts cannot, however, be attributed to this process as their compositions plot on a standard fractionation trend (Fig. 9).Rather, we hypothesise that the structural features of the Terceira Rift running roughly NW-SE through São Miguel's western half (Figure 1), which controlled the eruption of the mafic magmas in the Picos fissure system, may have played an important role.This relationship implies that suitable assimilants may occur in zones across the island as a whole (as suggested in other regions e.g.Iceland, Hampton et al. 2021), rather than being restricted to the central volcanoes where present-day hydrothermal activity is observed.

Potential assimilants
While some syenite lithics that were erupted within the Fogo A Plinian event are visibly hydrothermally altered and have low-δ 18 O values (2.6 to 5.6 ± 0.30‰ Snyder et al. 2004), the Ti/Zr vs K/Zr relations (as shown in Figure 9) rule out such evolved compositions from playing a significant role in petrogenesis of the Picos low-δ 18 O basalts.Rather, we suggest that the assimilant is likely to be altered oceanic crust (AOC), or other mafic lithologies within the island edifice.While the upper oceanic crust may have elevated δ 18 O compared to mantlederived melts (typically 7-15 ‰, Eiler 2001), values in the altered lower crustal

J o u r n a l P r e -p r o o f
Journal Pre-proof gabbros may reach values as low as 0 ‰ (Gregory and Taylor 1981;Muehlenbachs 1986;Staudigel et al. 1995).Other studies of ocean islands with low-δ 18 O mafic rocks have suggested a source within the volcanic edifice itself (e.g.Eiler et al. 1997;Thirlwall et al. 1997;Garcia et al. 1998) with recovered drillcore from the HGP-A well on the Big Island of Hawaii returning bulk δ 18 O values reaching as low as 3.2‰ (Garcia et al. 1998).Perhaps the most suitable potential assimilants are the rocks of altered layer 3 ocean crust (i.e.gabbros) from around the Canary Islands (Hansteen and Troll (2003), another mid-Atlantic ocean island chain whose AOC δ 18 O values range from 3.3-5.1 ‰.The occurrence of intrusive gabbroic bodies (8-12 km depth) is also reported for Pico Island in the Azores central group (Zanon et al. 2020), thus making it plausible that such rocks are present underneath São Miguel at similar depths.Taking a mid-point of the δ 18 O values range from the Canary Islands of 4.2 ‰ would imply a ~20% contribution from such lithologies to produce the lowest δ 18 O magmas found in the Picos fissure system using simple binary mixing (Sup.Fig. 6).This estimate is very much in line with that of Widom and Farquhar (2003) who suggested a hydrothermally altered component within the plume of 10-20%.The key difference with the interpretation of Widom and Farquhar (2003) is that our larger dataset highlights the spatial control on the low-δ 18 O mafic magmas and thus we prefer a shallow rather than deeper source for the assimilation.

Conclusions
This work represents the first detailed multi-mineral study of the oxygen isotopic composition of the mafic products of the island of São Miguel, Azores.From our new bulk rock, mineral, and isotopic data we can conclude: i) Analyses of olivine (4.58-5.36‰) and clinopyroxene (4.66-5.64‰) crystals indicate that these minerals are consistent with high-temperature equilibrium.
Our results indicate that low-δ 18 O mafic magmas exist on São Miguel (as suggested in previous studies) but the majority of analysed samples are similar to MORB.
ii) There exists little relationship between sample composition (e.g.MgO content, Ba, Zr, Ni, Th) and the δ 18 O values of mafic magmas on São Miguel.Nor is there a temporal relationship with both the highest and lowest-δ 18 O magmas erupted within the last few thousand years.
iii) A spatial relationship is however found, with the low-δ 18 O magmas concentrated in the Picos fissure system while those from the east (Nordeste) and west (Sete Cidades) sectors of São Miguel have higher values.This implies a local control on the δ 18 O of the mafic magmas rather than this being a deeper characteristic of the Azorean plume.
iv) The Picos fissure system contains mafic magmas that have assimilated abundant alkali feldspar from syenitic cumulates (Storey et al. 1989).The bulk compositions of our low-δ 18 O samples indicate that limited, if any, syenite cumulate involvement occurred and thus altered syenitic lithologies are unlikely J o u r n a l P r e -p r o o f Journal Pre-proof responsible for the low-δ 18 O values.Rather, we suggest that altered oceanic crust materials are assimilated to decrease the δ 18 O of Picos basalts.Figure 3 -Bulk rock compositions of São Miguel mafic rocks with our samples compared against literature data for São Miguel (Widom et al. 1996;Turner et al. 1997;Renzulli and Santi 2000;Beier et al. 2006Beier et al. , 2007;;Elliott et al. 2007;Prytulak et al. 2014 Moore (1990) and relative field stratigraphy.In cases of the Widom and Farquhar (2003) samples we have superimposed sample locations in their map to the geological map of Moore (1991) to try and estimate which unit was sampled.Where samples lie on a vertical line, we believe these samples are of the same unit, and the agreement of the data between the two studies supports this.Geochronology for the Nordeste samples remains imperfect but the best ages for this group of samples place them in the 878-780 ka bracket.Ellis et al. 2022).Notably the trachytes are lower in δ 18 O than the intercalated mafic rocks indicating that assimilation, and the lowering of δ 18 O, occurred during magmatic evolution likely within the crust.Reference trajectories for closed-system fractionation are shown from other oceanic island settings: G -Galapagos (Muehlenbachs and Byerly 1982.);T-Tristan da Cunha (Harris et al. 2000); A -Ascension (Sheppard and Harris 1985).

Figure captions
Figure 9 -Compositions of the samples in this study using a K/Zr and Ti/Zr plot (modified after Storey et al. 1989).This diagram shows that the low-δ 18 O samples in our study (highlighted) do not have elevated K/Zr as the hybrid lavas of the Picos fissure system.Thus the assimilation of syenitic or alkali feldspar cannot be used to explain the low-δ 18 O nature of these samples.

Declaration of interests
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Ben Ellis reports financial support was provided by Swiss National Science Foundation.

Figure 1 -
Figure1-Digital elevation model of São Miguel Island showing where samples in this study were taken from and identification of the volcanic systems (Sete Cidades volcano, Picos fissure system, Fogo volcano, also known as Água de Pau volcano, Congro fissure system, Furnas volcano, Povoação volcano, Nordeste volcanic system).Inset shows the location of the island of São Miguel (SM) within the Azores archipelago (StM -Santa Maria, T -Terceira, G -Graciosa, SJ -São Jorge, P -Pico, Fa -Faial, Fo -Flores, C -Corvo) in the North Atlantic and the main tectonic features of the region (Eu -Eurasian plate, Nu -Nubian plate, NA -North American plate, MAR -Mid-Atlantic Ridge, TR -Terceira Rift).
Figure3-Bulk rock compositions of São Miguel mafic rocks with our samples compared against literature data for São Miguel(Widom et al. 1996;Turner et al. 1997;Renzulli and Santi 2000;Beier et al. 2006Beier et al. , 2007;;Elliott et al. 2007;Prytulak et al. 2014) A. Total alkali vs. SiO2 plot, B. MgO vs Ni showing the occurrence of three samples in the Nordeste suite that have MgO values greater than 12 wt.%whichBeier et al. (2006) defined as accumulative in olivine.C. Ba vs Zr plot of two incompatible elements with the olivine accumulative samples having the lowest values D. Normalised REE plots using normalising data fromMcDonough and Sun (1995), showing similar trends between all samples from this study.The Nordeste accumulative samples show an overall decrease in REE due to the incorporation of olivine, but the pattern of the REE is the same as the other samples here.

Figure 4
Figure 4Olivine backscattered electron images and compositions.A) Zonation within single olivine crystals showing that some olivines from Picos and Nordeste are significantly zoned while those from Sete Cidades are homogeneous.B) Rhodes diagram showing few olivine compositions would be in equilibrium with their co-existing whole rock with similar results fromWidom and Farquhar (2003) shown in black and trajectories after Putirka (2008) C) Comparison of data from this study with those fromSobolev et al. (2005) shown in grey field.D) -G) Typical olivine appearance from the main regions studied here.

Figure 5 -
Figure 5 -Oxygen isotopes and magmatic evolution.A. Bulk rock composition (SiO2 wt.%) against δ 18 O values for all samples in this study illustrating expected behaviour following the summary of Troch et al. (2020).The grey box reflects the composition of MORB.B. Mineral-mineral plot of samples containing both olivine and clinopyroxene indicating the minerals are in high-temperature equilibrium.

Figure 7 -
Figure 7 -Diagram showing the lack of relationship between δ 18 O values and A. major elements -MgO and B. trace elements -Nb.

Figure 8 -
Figure8-O isotopic compositions of the youngest products of the Sete Cidades volcano (with trachyte data fromEllis et al. 2022).Notably the trachytes are lower in δ 18 O than the intercalated mafic rocks indicating that assimilation, and the lowering of δ 18 O, occurred during magmatic evolution likely within the crust.Reference trajectories for closed-system fractionation are shown from other oceanic island settings: G -Galapagos(Muehlenbachs and Byerly 1982.);T-Tristan da Cunha(Harris et al. 2000); A -Ascension(Sheppard and Harris 1985).

Table 1 -
All measured O isotopic compositions from minerals.