Magmatic evolution of the Kolumbo submarine volcano and its implication to seafloor massive sulfide formation

Seafloor massive sulfides form in various marine hydrothermal settings, particularly within volcanic arcs, where magmatic fluids may contribute to the metal budget of the hydrothermal system. In this study, we focus on the Kolumbo volcano, a submarine volcanic edifice in the central Hellenic Volcanic Arc hosting an active hydrothermal system. Diffuse sulfate-sulfide chimneys form a Zn-Pb massive sulfide mineralization with elevated As, Ag, Au, Hg, Sb, and Tl contents. These elements have similar behavior during magmatic degassing and are common in arc-related hydrothermal systems. Trace-element data of igneous magnetite, combined with whole rock geochemistry and numerical modelling, highlights the behavior of chalcophile and siderophile elements during magmatic differentiation. We report that, despite early magmatic sulfide saturation, chalcophile element contents in the magma do not decrease until water saturation and degassing has occurred. The conservation of chalcophile elements in the magma during magmatic differentiation suggests that most of the magmatic sulfides do not fractionate. By contrast, upon degassing, As, Ag, Au, Cu, Hg, Sb, Sn, Pb, and Zn become depleted in the magma, likely partitioning into the volatile phase, either from the melt or during sulfide oxidation by volatiles. After degas-sing, the residual chalcophile elements in the melt are incorporated into magnetite. Trace-element data of magnetite enables identifying sulfide saturation during magmatic differentiation and discrimination between pre-and post-degassing magnetite. Our study highlights how magmatic degassing contributes to the metal budget in magmatic-hydrothermal systems that form seafloor massive sulfides and shows that igneous magnetite geochemistry is a powerful tool for tracking metal-mobilizing processes during magmatic differentiation


Introduction
The study of modern ore forming systems is crucial when trying to understand the formation of ancient mineral deposits and provides useful models for the exploration industry.In this light, actively forming seafloor massive sulfide (SMS), considered analogous to ancient polymetallic volcanogenic massive sulfide (VMS) deposits, provides important insights into understanding the formation of the latter (Halbach et al. 1989;Hannington and Scott 1989).Seafloor massive sulfide can form anywhere on the seafloor where hydrothermal activity is present (i.e., at mid-oceanic ridges, submarine arc volcanoes, rifts, hot spots, e.g., Hannington et al. 2005).Arc-related hydrothermal systems differ in many aspects from those located along mid-oceanic ridges and mature back-arc spreading centers, mostly with respect to important magmatic contributions, including metals, to the hydrothermal system (e.g., de Ronde et al. 2001de Ronde et al. , 2003de Ronde et al. , 2005;;Hannington et al. 2005;Patten et al. 2020).
Magmatic volatile contributions to arc-related submarine hydrothermal systems like at the Kolumbo volcano (e.g., Carey et al. 2013;Kilias et al. 2013;Rizzo et al. 2016), Brothers volcano of the Kermadec arc (e.g., Berkenbosch et al. 2012;de Ronde et al. 2005de Ronde et al. , 2011;;Keith et al. 2018), and SuSu Knolls in the Manus back-arc basin (e.g., Craddock et al. 2010;Thal et al. 2016;Yeats et al. 2014) show similarities with on-land porphyry-epithermal deposits (de Ronde et al. 2019).In such porphyry-epithermal deposits, metals are considered to be mostly carried by magmatic fluids exsolved from hydrous magmas in the mid-to upper crust (Richards 2011).Thus, metal fractionation by magmatic processes such as sulfide saturation and degassing appear to be key parameters for controlling the formation of metal-rich magmatic fluids in arc-related hydrothermal systems (Du and Audétat 2020;Hedenquist and Lowenstern 1994;Jenner et al. 2010).Sulfide saturation in the magma will extract most of the chalcophile and siderophile elements of a silicate magma (including As, Ag, Au, Hg, Sb, and Tl) and concentrate them into sulfide phases-e.g., immiscible sulfide liquid or crystalline sulfides- (Fontboté et al. 2017;Jenner et al. 2010;Park et al. 2015;Sun et al. 2004).Similarly, magmatic degassing leads to formation of a ligand-rich aqueous phase (e.g., Cl, S, and F) able to form complexes with metals and remove them from the magma if it is still fertile-i.e., before sulfide saturation- (Audétat 2019;Fontboté et al. 2017).If the magma reaches sulfide saturation before volatile saturation, most of the chalcophile and siderophile metals partition into sulfide phases, preventing formation of metal-rich fluids upon magmatic degassing.However, volatile bubbles may nucleate on sulfide phases to form low-buoyancy sulfide-volatile compounds, transporting sulfides-and their metals-toward the shallow part of the magmatic system (Blanks et al. 2020;Mungall et al. 2015).In addition, oxidation of the sulfides by the volatiles within the sulfide-volatile compound may lead to metal transfer to the volatile phase and eventually to the magmatic-hydrothermal system (Patten et al. 2024).Therefore, investigating the timing between sulfide and volatile saturation and degassing in the melt is crucial to understand how these processes affect the metal fertility of the magma.To do so, petrogenetic indicator minerals, such as magnetite, have proven to be useful when constraining the geochemical conditions of a magmatic-hydrothermal system (Dare et al. 2014 and references therein).
Magnetite is a common mineral in magmatic-hydrothermal system and its composition may be used to identify its origin-i.e., hydrothermal, igneous, and/or ore-deposit related- (Dare et al. 2014;Huang et al. 2019;Nadoll et al. 2014).Trace-element compositions in igneous magnetite is mainly controlled by the temperature, fO 2 , and co-crystallizing phases; it provides insights into the melt composition at the time of magnetite crystallization and enables identification of simultaneous formation of other mineral phases, such as sulfides (Dare et al. 2014;Huang et al. 2019).Indeed, chalcophile elements are slightly incompatible to compatible in magnetite, and compositional variation of the melt may be recorded by the trace element content of magnetite (Dare et al. 2012).
In this study, we investigate the Kolumbo submarine volcano and its SMS in order to understand how magmatic processes contribute to the metal budget of SMS in volcanic arc magmatic-hydrothermal systems.Trace-element composition evolution of igneous magnetite and whole rock geochemistry is combined with petrology and numerical modelling to investigate the timing between sulfide saturation, volatile saturation, and degassing and to highlight how these processes affect the metal content of the magma during magmatic differentiation.This allows us to (1) estimate the metal fertility of the magma during magmatic differentiation, (2) determine the magmatic input of metals into the magmatic-hydrothermal system, and (3) ultimately derive the magmatic contributions to the formation of SMS.

Kolumbo volcano
The Kolumbo submarine volcano is located in the Anhydros sedimentary basin, southern Aegean Sea, as part of the Christiana-Santorini-Kolumbo volcanic field in the center of the 450-km-long Hellenic Volcanic Arc (HVA; Fig. 1 a-inset).Volcanic activity along the HVA is linked to subduction of the African Plate underneath the Aegean Microplate (Papanikolaou 2013), which can be traced back at least to the Pliocene (Fytikas et al. 1984;Preine et al. 2022).The HVA stretches from the Methana peninsula (Saronic Gulf) in the west through the islands of Milos, Antimilos, and Santorini (Cyclades) to Kos and Nisyros (Dodecanese) in the east, including submarine volcanoes (Nomikou et al. 2013;Fig. 1a-inset).Regional NW-SE extension in the west to NE-SW in the east leads to crustal thinning, exhumation of high-temperature metamorphic rocks in core complexes, and the deposition of extensional sedimentary basins (Agostini et al. 2010), including the Anhydros basin north-east of Santorini, between the islands of Ios and Anafi (Fig. 1a) (Nomikou et al. 2016).The Anhydros basin hosts the Kolumbo volcano as well as 25 other volcanic cones developed along the Christiana-Santorini-Kolumbo rift, forming the Kolumbo Volcanic Chain (Fig. 1a) (Hooft et al. 2017;Nomikou et al. 2012Nomikou et al. , 2019)).
The Kolumbo volcano has been the source of seismic activity over the past few decades (e.g., earthquake swarms of [2006][2007] and is considered the most active volcanic system in the region at the present time (Dimitriadis et al. 2010).Its volcanic cone has an elongated shape oriented NE-SW with a basal diameter of 3 km.The edifice rises from 500 m below sea level (mbsl) and shoals 19 mbsl at its shallowest point (Fig. 1b; Nomikou et al. 2012).The last eruption of 1650 CE formed a 1.7 km wide crater in the center of the cone (Fouqué 1879;Sigurdsson et al. 2006).The crater slopes are steep with the crater bottom at 502 mbsl (Nomikou et al. 2012).There is a hydrothermal vent field in the northern part of the crater hosting low-to high-temperature polymetallic Zn-Pb-(As, Ag, Au, Cu, Hg, Sb, Tl)-rich sulfate-sulfide chimneys (Fig. 1b; Carey et al. 2013;Kilias et al. 2013;Nomikou et al. 2022;Rizzo et al. 2016Rizzo et al. , 2019;;Sigurdsson et al. 2006).Lowtemperature chimneys emit clear fluids up to 70 °C and minor gas bubbles, whereas high temperature chimneys discharge CO 2 -rich fluids up to 265 °C (Nomikou et al. 2022;Sigurdsson et al. 2006).Mixing of magmatic fluid with the hydrothermal system at Kolumbo is estimated to occur at 270 °C and 1160 mbsl based on pressure and temperature geo-indicators (i.e., 650 m below the seafloor) (Rizzo et al. 2019).
Five distinct cycles of mostly submarine volcanism have been identified by seismic imaging (Hübscher et al. 2015;Preine et al. 2022).They form an assembly of five circular, cone-shaped stratigraphic units labelled K1 to K5 interbedded with volcano-sedimentary rocks (Fig. 1c) (Hübscher et al. 2015).The timing of volcanic activity is poorly constrained, but based on the stratigraphy, it is estimated to start with the second volcanic phase of the Christiana-Santorini-Kolumbo volcanic field between ca.1.6 and 0.7 Ma (Preine et al. 2022).Only the age of the K5 unit is known and linked to the 1650 CE eruption based on witness accounts (Hübscher et al. 2015;Richard 1657).
Kolumbo was sampled in 2010 by the remotely operated vehicle (ROV) Hercules during the R/V Nautilus cruise NA007 and later by drilling during the IODP (International Ocean Discovery Program) expedition 398.At the time of writing, only in situ samples from the outcropping units K2 and K5 collected by the ROV Hercules were available for study.The samples were collected from basaltic to andesitic lava flows and rhyolitic pumice layers (Klaver et al. 2016).Centimeter-sized mafic enclaves in pumice typically show chilled margins, indicating injection of mafic magma into a less hot, more evolved host magma (Cantner et al. 2014;Klaver et al. 2016).Klaver et al. (2016) suggest that the Al-rich nature of Kolumbo's magma results from wherlite cumulate formation in a lower crustal reservoir from a hydrous arc magma.Rhyolitic magma is produced by prolonged differentiation of this Al-rich melt, occasionally mixing with lower crustal melts (Klaver et al. 2016).It eventually rises adiabatically to the upper magma chamber until degassing occurs.Injection of the less evolved, hydrous, mafic magma at the base of the upper chamber likely acted as an eruption trigger (Cantner et al. 2014;Klaver et al. 2016;Konstantinou 2020).Seismic tomography and the analysis of seismic activity during the last 20 years indicates a shallow magma chamber (Chrapkiewicz et al. 2022;Dimitriadis et al. 2010;Schmid et al. 2022).Seismic tomography using data from earthquakes between 2002 and 2005 highlights a low velocity zone interpreted as a magma chamber at 6-7 km below the seafloor (Dimitriadis et al. 2010).Alternatively, Schmid et al. (2022) suggest that the strong seismic activity at these depths reflects a rheologically strong layer obstructing melt ascending toward a 2-4-km-deep magma chamber that was located by Chrapkiewicz et al. (2022) using seismic data from the PROTEUS experiment (Hooft et al. 2017). However, McVey et al. (2020), working on the same set of data, do not recognize any melts between 3 and 5 km depth.Although the exact depth and architecture of the actual uppermost magma storage at Kolumbo remains disputed, petrology-and geochemistry-based modelling by Cantner et al. (2014) estimates that rhyolitic magma was stored at a temperature of 750 °C at a depth of 5-6 km prior to the 1650 CE eruption.

Sample localities
The Hercules ROV allowed sampling of pumice and lava flow fragments from the K2 and K5 units, although with a sample size limit of ~ 15 cm because of limitations of the robotic arm (Fig. 1b).

Methods
Chemical analyses were performed at the Laboratory for Environmental and Raw Materials Analysis at the Institute of Applied Geosciences in Karlsruhe, Germany, with the exception of electron probe micro-analysis (EPMA), which was performed at the Institute of Geological Sciences of the University of Bern, Switzerland, and Hg content measurement by atomic absorption spectroscopy at the Oceanic and Continental Environment and Paleoenvironment mixed research unit (UMR EPOC) in Bordeaux, France.Petrography was done via transmitted and reflected light microscopy.

Whole rock geochemistry
Major element compositions of whole rocks (n = 13) were measured by X-ray fluorescence (XRF; S4 Explorer, Bruker AXS) on glass beads with BHVO-1, MRG-1, RGM-1, and SY-2 as reference materials, as a measure of analytical precision and accuracy (Table 1, ESM 1).Carbon and sulfur contents were measured on rock powder by solid state infrared absorption using a Carbon-Sulfur Analyzer (CS-2000, Eltra) with steel (92,400-3050), ductile iron (92,400-3100), and barium sulfate (90,821) standards from Eltra as reference material as a measure of analytical precision and accuracy (see ESM 1).Trace element contents (n = 13, Table 2) were measured by laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) on pressed powder pellets using a Teledyne 193 nm Excimer Laser coupled to an ICP-MS (Element XR ThermoFisher) with spot size of 85 µm; laser frequency of 10 Hz, fluence of 5 J•cm −2 ; and He, Ar, and N flow of 0.3 L•min −1 , 0.85 L•min −1 , and 10 mL•min −1 , respectively.Pressed powder pellets were prepared following the method described in Patten et al. (2023).Calibration and data quality were checked using pressed powder pellets of standards BHVO-1, BHVO-2, BCR-2, and BIR-1 from the USGS.In order to improve data quality, Au contents were measured a second time, on the same equipment, following the method of ultra-low detection of Au on pressed powder pellets by LA-ICP-MS developed by Patten et al. (2023).Calibration and data quality were checked using pressed power pellets of standards BHVO-2, BCR-2, and BIR-1 from the USGS, MRG-1, and TDB-1 from NRCan and TSD-41, an in-house epidosite standard.Data reduction of all LA-ICP-MS analysis was done using the Iolite software 3DRS plugin v.4.8.3 (Paton et al. 2011).Accuracy and precision for reference materials (< 15% for most elements) as well as limits of detections are detailed in Supplementary data (see ESM 1).The Hg content was measured by spectrophotometry with a direct Hg analyzer DMA-80 (Milestone) using standards BCR-277R and TCEGir (in-house) for calibration (see ESM 1).

Magnetite composition
The major element compositions of magnetite (n = 77 from 7 samples, Table 3) were measured by EPMA (JEOL JXA-8200 Superprobe) using the following synthetic and natural standard reference materials: anorthite (Al 2 O 3 , CaO), orthoclase (SiO 2 ), magnetite (FeO), forsterite (MgO), rutile (TiO 2 ), metal nickel (NiO), spinel (Cr 2 O 3 ), sphalerite (ZnO), and pyrolusite (MnO).Spot analyses for Al, Si, Ca, Fe, Mg, Ti, Ni, Cr, Zn, and Mn were performed using 15 keV accelerating voltage, 20 nA specimen current, and 40 s dwell time (10 s for each background after 20 s on a peak) (see ESM 2).The trace element compositions of magnetite (n = 143 from 8 samples, Table 4) were measured by in situ LA-ICP-MS analysis, on the same equipment as for the pellet LA-ICP-MS analyses, with spot size of 35 µm, laser frequency of 10 Hz, fluence of 5 J.cm −2 , and He and N flow of 0.3 L•min −1 and 10 mL•min −1 , respectively.Iron data from the EMPA was used for internal standard calibration of the LA-ICP-MS data.Calibration and data quality was checked using Fe-rich basaltic glasses BHVO-2, BCR-2, and BIR-1 from the USGS, following the method from Dare et al.

Petrography
Samples recovered from the Kolumbo volcano come from the K2 and K5 volcanic units (n = 3 and n = 10, respectively).Two types of samples were collected: intermediate to felsic pumice (K2 and K5 units) and mafic to felsic lava (K5 unit only; Fig. 2a-d).As the samples are mostly composed of a glass to microcrystalline matrix, their igneous classification cannot be determined petrographically.Further distinctions (e.g., rhyolite pumice vs. trachyte pumice) are based on geochemical data described below.
The enclaves have a quenched texture on their margin where the contact with the matrix is sharp, indicative of rapid crystallization (Fig. 2c).The contact usually displays increased vesicularity when compared to the rest of the rock.

Magnetite
Magnetite is ubiquitous within the volcanic samples of Kolumbo and shows three main habits: (1) small (< 50 µm), idiomorphic magnetite in enclaves where it is disseminated in the matrix and forms inclusions in silicates or is locally associated with sulfides (Fig. 2e); (2) sub-idiomorphic to idiomorphic magnetite (< 100 µm) in the matrix, where it is usually fractured or fragmented and locally contains sulfide inclusions (Fig. 2f); and (3) subidiomorphic magnetite (< 200 µm) in clusters, where it is usually fractured and rounded, locally has sulfide inclusions, and may show compositional zoning (Fig. 2f, g).

Sulfides
Sulfides, although relatively scarce, are present in most of the volcanic samples of Kolumbo.The most common mineral is pyrrhotite along with lesser chalcopyrite and pyrite (Fig. 2e, f, i-k).They form xenomorphic grains that are usually < 100 µm, but can be at the mm-scale (Fig. 2j,  k).The grains are usually monomineralic (i.e., pyrrhotite or chalcopyrite) but may form a pyrrhotite-chalcopyritepyrite assemblage, locally replaced by magnetite and goethite (Fig. 2i).In addition, sulfides locally occur as inclusions (< 10 µm) in all types of magnetite and in the phenocrysts (Fig. 2f, h).

Whole rock geochemistry of pumice, lava, and enclave samples
Major element compositions of pumice, lava, and enclave samples indicate a lithological diversity ranging from   basaltic andesite to rhyolite where the trace element content varies as the MgO content decreases (Table 1 and 2).

Magnetite geochemistry
Magnetite shows limited major element variation in all samples (Table 3).The Al, Co, Cr, Cu, Ga, Mg, Ni, Sc, and V contents are higher in magnetite hosted in more mafic samples than in those hosted in felsic samples.Conversely, the Ag, As, Bi, Ge, Hf, Mn, Mo, Nb, P, Pb, Sb, Sn, Ta, Ti, W, Y, Zn, and Zr contents are higher in magnetite from felsic samples than in those in more mafic samples (Table 4).

Melt evolution modelling with COMAGMAT 3.75
The evolution of the melt was modelled using COMAG-MAT 3.75 (Ariskin and Barmina 2004) to decipher the processes leading to metal mobilization from the melt.This model is based on a combination of empirical and thermodynamic calculations of mineral-melt equilibrium and can simulate magma fractionation.An average whole rock composition of the three most mafic sample compositions was used as a proxy for the starting composition of the melt and the accuracy of the model is controlled by geochemical data.However, because there was a limited number of available samples, the model is not very well constrained between 0.8-2.0 and 2.5-3.0 wt% MgO.The pressure constraints were set to decrease from 200 to 90 MPa to reflect the ascending melt in an upper reservoir located between 6 and 3 km depth.Multiple runs of the model using various redox conditions (fO 2 FMQ + 0.5 to 1.5) and variable H 2 O content (1 to 3 wt%) show that the starting conditions producing a model best fitting the evolution of the major element compositions are fO 2 FMQ + 1 and 2.0 wt% H 2 O (Fig. 3).Using these best fit parameters, the model suggests that magnetite crystallization starts as  the melt reaches 2.5 wt% MgO, whereas the petrography shows an earlier occurrence of magnetite.This discrepancy is addressed in the following discussion.In addition, the water content increases during magmatic differentiation from 2.0 to 4.3 wt% H 2 O at 0.9 wt% MgO, whereupon water saturation is reached and the water content decreases to 2.9 wt% in the most evolved melt.

Magmatic differentiation at Kolumbo
Major element compositions of our samples together with data from previous studies (Cantner et al. 2014;Klaver et al. 2016) show that the K2 and K5 units record similar differentiation processes, i.e., from basaltic andesite to rhyolite.According to Klaver et al. (2016), they reflect the progressive magmatic differentiation of an Al-rich hydrous mafic melt from a lower crustal reservoir to the upper magma chamber.

Petrology and geochemistry
The petrological and geochemical diversity of the samples illustrates a complex magmatic history.Mafic enclaves with chilled margins in intermediate-to-felsic rocks is characteristic of injection of mafic magma into an evolved magma chamber (Fig. 2b-d), whereas samples with a trachytic composition (e.g., NA007-002c) form from magmatic differentiation of the mafic magma or mixing with the evolved magma.The basaltic andesite or andesite composition of the mafic enclaves indicates that their parent melt is not a direct product of mantle melting, but already has started to differentiate in a deeper magma chamber before reaching the shallower, evolved magma chamber.This is consistent with the compatible element-depleted nature of the most mafic rocks (as observed in the whole rock trace element data), likely a result of fractionate magmatic differentiation (Fig. 4).Slightly lower Ti and V relative to mid-ocean ridge basalt (MORB) may be the result of early magnetite crystallization, whereas lower Au, Co, Cu, and Ni contents likely reflect early sulfide formation.Partitioning of Ni ± Co in olivine (Spandler and O'Neill 2010) throughout wherlite formation during the early magmatic differentiation history (Klaver et al. 2016) may also account for their higher depletion in basaltic andesite in regard to other siderophile and chalcophile elements (Fig. 4).The high As, Ag, Sb, Tl, and Pb contents in the basaltic andesite relative to MORB may either be related to early differentiation, as these elements are mostly incompatible (Fig. 4), or to mantle enrichment due to supra-subduction zone contamination by slab dehydration, particularly for As, Sb, and Pb (Hattori et al. 2005;Patten et al. 2017).

Melt evolution modelling
The best fit parameters of fO 2 FMQ + 1 and 2 wt% H 2 O are within the range given for other arc volcanic systems (i.e., ΔFMQ ≈ + 1 ± 1) where the mantle source is relatively more oxidized than in mid-ocean ridge setting (Bucholz and Kelemen 2019;Richards 2015).The modelled evolution of the SiO 2 , TiO 2 , Al 2 O 3 , CaO, and K 2 O contents with decreasing MgO, or increasing differentiation, fit well with sample compositions (Fig. 3).The modelled FeO content is slightly higher as measured in the samples during the early differentiation stages (> 2 wt% MgO), possibly due to model limitations preventing simulation of early magnetite formation (Fig. 3).The modelled P 2 O 5 content increases with progressive differentiation but decreases in the geochemical data (Fig. 3).This discrepancy is caused by apatite crystallization and fractionation that is not supported by the modelling.Apatite crystallization is contemporaneous with magnetite formation (Fig. 2e), leading to decreasing P 2 O 5 contents in the melt as magnetite crystallization increases (cf.Jenner et al. 2010).The Na content in the modelled melt reproduces the sample composition during differentiation until 0.5 wt% MgO, then it decreases as albite forms.Conversely, Na contents remain constant in the samples with MgO < 0.5 wt%, indicating that plagioclase did not fractionate during late differentiation.Except for the minor limitations discussed above, the model describes reasonably well the sample geochemistry.The key pieces of information that can be drawn from the model is increased magnetite formation below 2.5 wt% MgO and water saturation at ~ 0.9 wt% MgO.

Continuous magnetite crystallization or magnetite crisis?
The magnetite crisis, as defined by Jenner et al. (2010), refers to chemical changes in the melt related to sudden magnetite crystallization during melt differentiation at ~ 2.5 wt% MgO and is a critical mechanism to mobilize chalcophile elements during arc magma evolution.Formation of magnetite drives redox in the melt toward reduced conditions by incorporating two atoms of Fe 3+ for one of Fe 2+ in its crystal lattice.This favors sulfide saturation and chalcophile element partitioning from the silicate melt into sulfide phases (Jenner et al. 2010).Magnetite crisis is generally expressed in the melt chemistry by a sharp drop in FeO 2 , TiO 2 , and V/Sc.Limitations of the model preventing magnetite formation at MgO > 2.5 wt% cause modelled FeO 2 and TiO 2 contents and V/Sc values to increase during magmatic differentiation (Fig. 3).Whole rock data, however, display continuously decreasing FeO 2 and TiO 2 contents and V/Sc values with magmatic differentiation until 2.5 wt% MgO, indicating early igneous magnetite formation, in accordance with petrographic observation (Figs.2e, 3, and 5a).Below 2.5 wt% MgO, FeO 2 and TiO 2 contents decrease strongly, reflecting increased magnetite formation, as supported by modelling (Fig. 3).In addition, a drop in the V/Sc value is a good proxy for magnetite formation because V is more compatible in magnetite than Sc (Dare et al. 2014;Jenner et al. 2010).Early magnetite will have a higher V/Sc value than magnetite forming at a later stage of magmatic differentiation.Decreasing V/Sc values in igneous magnetite supports magnetite fractionation upon magmatic differentiation and suggests that magnetite reflects the composition of the melt in which it crystallizes (Fig. 5b).
For magnetite hosted in rhyolite, a few grains have higher V/Sc values indicating that these are exogenous and formed from a more mafic melt (Fig. 5b).In summary, it seems that Kolumbo's magma undergoes continuous magnetite crystallization from the early stages of its differentiation history, with the rate increasing notably below 2.5 wt% MgO.

Sulfide saturation
Sulfur concentration at sulfide saturation (SCSS) during magmatic differentiation is modelled using the method developed by Smythe et al. (2017) with a melt composition calculated by COMAGMAT 3.75.In comparison to SCSS values modelled for MORB (cf.Jugo 2009; SCSS = 2000 µg•g −1 S), the SCSS in Kolumbo's most primitive melts is relatively low (~ 300 µg•g −1 S) and indicates that the melt is sulfide-saturated during the entire differentiation process (Fig. 5c).The presence of sulfides in all samples supports sulfur saturation throughout the differentiation process and their habit (i.e., droplets and blebs) indicates crystallization from a sulfide liquid (Fig. 2e-k) (Patten et al. 2013).Samples with compositions between 2.0 and 4.0 wt% MgO have S values below the SCSS (Fig. 5c) despite the presence of sulfide droplets in the matrix and inclusions in phenocrysts (Fig. 2e, i, k).This apparent low S content is likely the result of S loss by degassing upon eruption, while outlying high S values (i.e., > 500 µg•g −1 ) in felsic samples are explained by seawater/rock interaction upon eruption (Brandl et al. 2022).The mechanisms triggering early sulfide saturation are unclear but may be related to magnetite or amphibole crystallization (Georgatou et al. 2022;Jenner et al. 2010).Similarly, amphibole replacing clinopyroxene modifies the redox state of the melt and may trigger sulfide saturation as the primitive basaltic melt evolves toward an andesite composition in the lower crustal reservoir (Georgatou et al. 2021(Georgatou et al. , 2022)).This process is however not clearly identified in Kolumbo.

Degassing
Understanding the degassing processes at Kolumbo is critical for understanding metal transfer in the magmatichydrothermal system given that degassing volatiles can carry and transport ligands and metals (Hedenquist and Lowenstern 1994).However, the initial volatile content of the samples may be altered by eruption degassing and interaction with seawater (Fig. 5c).Thus, the best proxy to assess degassing is to use water saturation indicated by the modelling of the melt evolution (i.e., 0.9 wt% MgO).However, initial volatile content in the model (2 wt% H 2 O) is lower than the average loss on ignition measured (3.2 wt%) in the most mafic samples (Table 1).This is likely caused by seawater/rock interaction after eruption increasing the volatile content in the samples in regard to their pre-eruption values (Coogan and Gillis 2018;Haraguchi et al. 2014).Loss on ignition values for rhyolite yield a post-water saturation volatile content of 2.8 wt% in lava, which is nicely reproduced by the model but increasing up to 7.1 wt% in pumices.However, as pumice forms during the eruption as a mixture of degassing volatiles and magma, it may not be representative of the initial volatile content.Degassing of magmatic volatile species is controlled by the confining pressure, temperature, oxidation state, and volatile content in the magma.Usually, magmatic volatiles are released in the following order: CO 2 < SO 2 -H 2 S < H 2 O (and halogens), according to their solubility in the melt (Mavrogenes and O'Neill 1999;Rouwet et al. 2019).As CO 2 is saturated early in magmatic arc systems, it starts to separate from the silicate melt in the lower crust (Lowenstern 2001).In the case of Kolumbo, CO 2 effervescence likely occurs in the lower crust where differentiation from basalt to basaltic andesite occurs.
Our model predicts water saturation at 0.9 wt% MgO (i.e., during trachyte formation) in the upper magma chamber (between 7 and 2 km depth).This indicates that SO 2 degassing could have taken place during magma ascend, or in the upper magma chamber.However, early sulfide saturation might have prevented SO 2 degassing by sequestrating S in a solid or liquid sulfide phase.As the magma degasses, volatile bubbles can nucleate on sulfide phases, leading to the formation of sulfidevolatile compounds (Barnes et al. 2019;Mungall et al. 2015).

Pre-water saturation magmatic differentiation
Whole rock data indicate that before water saturation is reached during magmatic differentiation, V, Ni, Co, Cu, Ti, Au, and Zn are progressively depleted, whereas Sn, Ag, As, Sb, Pb, and Tl are enriched (Fig. 6a-f; Table 2).Depletion in V, Ni, Co, and to some extent Fe and Ti is related to magnetite crystallization and settling (c.f.Dare et al. 2014).Despite early sulfide saturation, there is no notable chalcophile element depletion in the magma before rhyolite formation and water saturation, implying more complex processes during the magmatic evolution than simple sulfide phase formation and settling (Fig. 6a-d, Table 2).

Insight from magnetite before water saturation
Magnetite in Kolumbo's volcanic rocks is of magmatic origin and almost exclusively primary (Fig. 5a, b).Thus, magnetite trace element compositions reflect the metal content in the melt, whereas the whole rock composition is considered representative of the magma (i.e., melt plus mineral phases) composition.The contrasting metal content between magnetite and its host rock provides first-order insights into metal mobilizing processes during magmatic Fig. 4 Whole rock trace element data normalized to average mid-oceanic ridge basalt, based on Arevalo and McDonough (2010) differentiation (Table 4).From basaltic andesite through trachyte differentiation stages, the Cu content in the magma shows a slightly decreasing trend but remains almost constant (avg.28 µg•g −1 ) while it decreases from an average of 141 to 10 µg•g −1 in the corresponding magnetite (Fig. 6a).While Cu remains in the magma, its availability in the melt for magnetite decreases during magmatic differentiation.This indicates that the majority of Cu is sequestered in mineral phases, likely sulfide phases that predominantly stay in the magma.Nickel and Co show a different behavior.Nickel is depleted simultaneously in magnetite and in the magma during magmatic differentiation (Fig. 6e).This indicates that, unlike any sulfide phases, magnetite fractionates from the magma after crystallizing (Fig. 7a stage 1).Cobalt is less compatible than Ni in magnetite and is consequently slightly less depleted in the magma as magnetite fractionates (Fig. 6f) (Dare et al. 2014).

Metal budget at sulfide saturation
At the Kolumbo volcano, the melt is sulfide-saturated at least from the basaltic andesite differentiation stage on, as observed petrographically and supported by the modelling (Figs.2e-k and 5c).A consistent drop in the Cu/Ag value during differentiation indicates monosulfide solid solution (MSS) formation from a sulfide liquid or precipitation of Cu-rich sulfide (Figs.2e-k and 8) (Du and Audétat 2020;Jenner et al. 2010Jenner et al. , 2015)).Despite early sulfide saturation, chalcophile elements in the whole rocks are not significantly depleted during magmatic differentiation and are even enriched for Ag, As, Au, Bi, Cu, Hg, Pb, Sb, Sn, and Tl prior to water saturation, indicating that most sulfides phases do not settle and remain in suspension in the magma during differentiation (Figs.6a-d and 7a stage 1; Table 2).This is likely linked to the increasing melt viscosity as it differentiates and limited size of the sulfide phases (i.e., < 10 µm for disseminated sulfides; Fig. 2k) which both drastically reduce the settling velocity (< 1 cm.year −1 for micrometer-scale sulfides; see Holzheid 2010).Settling of sulfides phases can also be prevented by the formation of sulfide-volatile compounds as the magma degasses (Mungall et al. 2015;Yao and Mungall 2020).That is, as a volatile bubble grows on a sulfide liquid droplet, the average buoyancy of the compound can eventually lead it to "float" in the melt or to rise upward (Yao and Mungall 2020).

Metal mobility upon degassing
Magmatic degassing is an efficient process to mobilize metals from a magmatic melt and transport them as sulfide or chloride complexes (Gammons and Williams-Jones 1997;Nadeau et al. 2021).At Kolumbo, according to modelling, water saturation is reached at 0.9 wt% MgO during the trachyte differentiation stage.Previous degassing that could have occurred during magmatic differentiation (i.e., of CO 2, SO 2 , H 2 S) did not significantly deplete chalcophile elements in the magma prior to water saturation (Fig. 6a-d; Table 2).
A first-order insight of metal mobility upon volatile degassing is given by comparing the metal content in the "degassed" samples (i.e., after water saturation) with "undegassed magma."The evolution of metal content in "undegassed magma" is extrapolated for compositions < 0.9 wt% MgO using a curve fitting pre-water saturation data (i.e., > 0.9 wt% MgO).It appears that the most chalcophile elements such as Ag, Au, Cu, and Hg (with the notable exception of Bi and Tl), as well as less chalcophile elements like As, Pb, Sb, Sn, and Zn, are depleted in the samples after water saturation in comparison to the "un-degassed magma" (Figs.6a-d and 7a stage 2, b).Chalcophile elements are mostly enriched in sulfide phases prior to water saturation, and therefore, their depletion in the "degassed" samples implies metal transfer between the sulfide phases and the volatiles.Although Ni and Co are not depleted in comparison to "un-degassed magma," their content in the magma continues to decrease after water saturation (Fig. 6e,  f).The progressive loss of Ni and Co is probably controlled by continuous magnetite crystallization and fractionation (Fig. 7a stage 2, b).Formation of sulfide-volatile compounds may lead to the oxidation and resorption of the sulfide phase, transferring the metals to the volatiles and contributing to the formation of a metal-rich fluid (Edmonds and Mather 2017;Mungall et al. 2015).

Magnetite as metal flux tracer
The different compatibility of chalcophile elements in sulfides, volatile phases, and magnetite allows tracking metal mobilization during magmatic differentiation.Given that igneous magnetite at Kolumbo is primary, the V/Sc value of the magnetite allows tracking of the degree of magmatic differentiation and can be calibrated with the whole rock compositions of the host rock (Fig. 9).Decreasing Cu contents in magnetite during magmatic differentiation and before water saturation reflects sulfide saturation (Fig. 9).High Cu contents but low V/Sc values are characteristic of post-degassing magnetite that formed after water saturation (Fig. 9).Following degassing, chalcophile elements are locally enriched within magnetite (Figs. 2g and 6a,b).This behavior is the result of the competing partitioning of chalcophile elements between sulfides and magnetite.With S absent from the melt after degassing, remaining chalcophile elements that did not partition into volatiles are then incorporated into the magnetite (Fig. 7a stage 3, b).

Comparison with other arc volcanoes
Sulfide saturation, formation, and fractionation of sulfide phases and volatile degassing are among the main processes controlling chalcophile element contents in the melt during magmatic differentiation (Edmonds and Mather 2017;Jenner et al. 2010Jenner et al. , 2015;;Patten et al. 2013).The Fig. 6 Chalcophile and siderophile element contents of rocks and magnetite during differentiation as shown by whole rock MgO data.The left and right axes are for the whole rock and magnetite data, respectively.Note: scale of whole rock and magnetite data are different.The blue dotted lines connect the medians of the magnetite data.The vertical blue lines represent water saturation at 0.9 wt% MgO according to modelling.Fitting curves (in gray) are calculated based on data > 0.9 wt% MgO (i.e., before water saturation) and indicate the evolution of metal content in the whole rock during magmatic differentiation.The dotted segments of the curves are extrapolated toward MgO = 0 wt% to simulate the evolution of metal content in the whole rock during magmatic differentiation, in the absence of water saturation, degassing, and associated metal removals whole rock Cu/Ag value compared to the MgO content illustrates how chalcophile elements are mobilized, highlighting whether or not a sulfide liquid or crystalline sulfide phase forms at sulfide saturation (Du and Audétat 2020;Jenner et al. 2010;Li and Audétat 2012;Patten et al. 2013).Formation of a sulfide liquid does not change the Cu/Ag value, as both elements have similar partition coefficient in sulfide liquid (Li and Audétat 2012;Patten et al. 2013).Conversely, crystallization of sulfides (i.e., MSS or Cu-rich sulfides) leads to a decrease of the Cu/ Ag value because Cu is more compatible than Ag in these phases (Du and Audétat 2020;Jenner et al. 2015;Li and Audétat 2012).In a basalt, for example, the Cu/Ag value is usually close to the mantle value of ~ 3000 ± 500 and remains unchanged during magmatic differentiation, provided only sulfide liquid forms (e.g., at Niuatahi volcano; Fig. 8) (Brandl et al. 2022;Park et al. 2015;Wang et al. 2019).Crystallization and fractionation of sulfides lead to a progressively decreasing Cu/Ag value from the mantle value in the whole rock during magmatic differentiation, as observed in samples from the Manus and Lau basin of the Valu Fa Ridge (Tonga), Brothers volcano, and adjacent ridges of the Kermadec arc (Fig. 8) (Brandl et al. 2022;Jenner et al. 2010Jenner et al. , 2012Jenner et al. , 2015)).
The Cu/Ag values of Kolumbo samples reflects a differentiation process similar to the early differentiation stages of the Nisyros volcano (Georgatou et al. 2022).The Cu/Ag values at Kolumbo are interpreted to yield mantle-like ratio at basaltic composition and decrease continuously until ~ 0.9 wt% MgO, indicating uninterrupted formation of magmatic sulfides since early differentiation in the lower crust until the trachyte/rhyolite stage in the upper magma chamber (Fig. 8).Degassing at ~ 0.9 wt% MgO leads to a sharper drop of the Cu/Ag value, Fig. 7 Conceptual model of metal mobility during magmatic differentiation and related hydrothermal processes.a 1.Before degassing, the melt is S-saturated, chalcophile elements (i.e., Ag, As, Au, Cu, Hg, Sn, Sb, Pb, and Zn) fractionate preferably into sulfide phases while Ni, Co, V, and Ti are partitioned preferably into magnetite.The sulfide phases are "floating" in the magma while magnetite settles.2. Upon degassing, the sulfide phases are oxidized by the formation of sulfide-volatile compounds or interaction with volatiles.Chalcophile elements are removed from the sulfides and likely transported as Sor Cl-complexes.3.After degassing, the system is S-depleted and the remaining chalcophile elements in the melt partition into magnetite, locally forming chalcophile-element-rich magnetite.b Chalcophile and magnetite-compatible elements in whole rock and magnetite during differentiation.c Model of metal transfer within Kolumbo's magmatic-hydrothermal system indicating mobilization of Cu by magmatic fluids as Cu is more compatible than Ag in a volatile phase exsolving from the melt (Fig. 8) (Audétat and Edmonds 2020).

Metal sources
Deep hydrothermal fluid circulation in the oceanic crust is an efficient process for mobilizing metals from the country rocks and partially controls the metal endowment in SMS/ VMS deposits (Hannington et al. 2005).In mid-oceanic ridge settings, the oceanic crust is thought to represent the source of metals (i.e., As, Au, Cu, Pb, Se, Sb, and Zn) that are mobilized by hydrothermal leaching (Brauhart et al. 2001;Patten et al. 2016;Skirrow and Franklin 1994).Similarly, in volcanic arc environments, especially in continental crust, felsic country rocks are enriched in Ag, As, Pb, Sb, Sn, and Tl compared to MORB (Fig. 4), and may be a source for these elements that are commonly enriched in related SMS/VMS deposits (Hannington et al. 2005;Shikazono 2003;Shu et al. 2022;Stanton 1994).At Kolumbo, the sulfide-sulfates chimneys are Zn-Pb-rich with relatively high As, Ag, Au, Hg, Sb, and Tl contents, but are relatively Cu poor (Kilias et al. 2013).Recent study of δ 65 Cu in sulfide minerals of the chimneys supports significant magmatic contributions (i.e., δ 65 Cu range of ∼ 0‰) and strongly suggests that direct magmatic degassing supplies volatiles and Cu to the hydrothermal system at Kolumbo (Zegkinoglou et al. 2023).These findings are concordant with Cu depletion during magmatic differentiation, interpreted as loss during volatile degassing and suggest that As, Ag, Au, Hg, Pb, Sb, Sn, and Zn are mobilized in the same way (Fig. 10).Conversely, Tl is enriched in Kolumbo's chimneys but does not seem to be mobilized during degassing, which suggests that metal sources other than the magma are involved in the magmatic-hydrothermal system (Figs.6d and 10).Similar metal enrichment has been linked to magmatic contribution to hydrothermal systems in other volcanic arc SMS, especially at the extensively studied Brothers volcano (Berkenbosch et al. 2015(Berkenbosch et al. , 2019;;de Ronde et al. 2011de Ronde et al. , 2019)).In Brothers chimneys, high enrichment factors of Ag, Au, and Cu with regard to the host rock indicate a magmatic source, whereas those of As, Pb, Sb, Sn, Tl, and Zn are closer to host rock values, pointing toward a mixed magmatic and host rock source (Berkenbosch et al. 2019).Thus, the metal association at Kolumbo likely results from a combination of hydrothermal wall rock leaching and magmatic input-i.e., magmatic volatiles and/or hypersaline brines (see de Ronde et al. 2019;Diehl et al. 2020)-(Fig. 7c, 10).Constraining metal sources is fundamental to understand the formation of SMS, but it is crucial to also consider the magmatic-hydrothermal processes that may affect the metal content in the mineralizing fluid and ultimately the metal association in the chimneys.

Magmatic-hydrothermal control on metal association
Submarine arc volcanoes host complex and dynamic magmatic-hydrothermal systems with constantly evolving fluid chemistry through space and time (Diehl et al. 2020;Lee et al. 2023;Lilley et al. 2003).Magmatic-hydrothermal processes occurring on and below the seafloor (e.g., phase separation, transient magmatic input, and mixing with brines) may drastically affect the metal content of the fluids (Diehl et al. 2020;Hannington 2014).Despite Cu being mobilized during degassing, it is not strongly enriched in the hydrothermal chimneys at Kolumbo, suggesting that Cu may be trapped in the deeper/hotter parts of the hydrothermal system (Fig. 10).We identify four mechanisms that may impact the metal content in the mineralizing fluid: cooling, boiling, phase separation, and fluid mixing.et al. 2019).Under these conditions, the magmatichydrothermal fluid is very close to boiling (Bischoff and Rosenbauer 1984) and eventually does as it ascends toward the seafloor.This is supported by the extreme Cu isotope fractionation in the minerals of the chimneys (Zegkinoglou et al. 2023).Boiling-induced precipitation of chalcopyrite and other sulfides at depth may explain the lack of Cu and Sn observed in the chimneys, despite mobilization during magmatic degassing (Figs.6a and 10; Table 2).3) Magmatic degassing generates a single-phase liquid of relatively low salinity (2-10 wt% NaCl eq.), which tends to separate into a hypersaline brine (40 to > 60 wt% NaCl eq.) and a vapor phase as it decompresses (< 4 km depth) and cools (Sillitoe 2010 and references therein).Phase separation causes metal fractionation with Zn, Ag, Sn, Pb, and Tl partitioning into the brine and As, Au, and Cu into the vapor (Heinrich et al. 1999).In porphyry Cu systems, Cu sulfides precipitate upon fluid cooling over the 425 to 350 °C range, assisted by fluidrock interaction (Sillitoe 2010 and references therein).Similar processes at Kolumbo may lead to formation of Cu porphyry-style mineralization at depth, which later remobilization by magmatic-hydrothermal fluids could provide metals to the chimneys (Martin et al. 2023).4) Submarine magmatic-hydrothermal systems are complex systems hosting various fluids of diverse origins: i.e., seawater-eventually modified by fluid-rock interaction-connate brines, magmatic-derived vapor, and brines (Diehl et al. 2020;Hannington et al. 2005).Magmatic brines can form lenses and be stored in the permeable sub-surface for hundreds of thousands of years Please note that the vertical scale is increased below 1. Nickel and Co are not shown here because of their siderophile nature and strong partitioning in magnetite, causing depletion between trachyte and rhyolite that is not related to degassing (Afanasyev et al. 2018).Changes in the hydrothermal fluid regime may lead to interaction of seawater with these metal-rich fluids, providing metals for the formation of chimneys on the seafloor (de Ronde et al. 2019).
Lenses of fluids trapped in the volcaniclastic sequence at Kolumbo have been identified (Hübscher et al. 2015) and may represent magmatic brines.Mixing of such fluids with hydrothermal fluids could contribute to the Ag, Pb, and Zn enrichments observed in the chimneys, as these metals preferentially partition into brine during magmatic fluid phase separation (Heinrich et al. 1999).
The metal association observed in chimneys at Kolumbo corresponds to low-temperature VMS deposits (< 300 °C), which usually show enrichment in Zn, Pb, Ag, Au, Cd, Sn, Sb, As, Hg, ± Tl, and ± W (Hannington 2014).The high contents of As, Sb, Hg, and Tl in the chimneys support boiling and zone refining processes, as these elements are easily mobilized in the vapor phase (Hannington 2014).However, the metal content of the chimneys likely does not reflect a straight forward process as the evolving dynamic nature of submarine magmatic-hydrothermal systems could include periods of magmatic quiescence that results in a seawater-dominated hydrothermal fluid, whereas periods of intense magmatic activity would increase the proportion of magmatic fluids (Diehl et al. 2020).Tectonic or magmatic events, changes in permeability may affect the dynamics of the hydrothermal fluid regime, leading to mobilization of different fluids and ultimately impacting the metal budget (de Ronde et al. 2019).Finally, eventual mineralization caused by the different processes discussed above may be overprinted by later hydrothermal fluid circulation, eventually remobilizing the metals toward the seafloor (Martin et al. 2023).

Conclusions
The Kolumbo submarine arc volcano and its SMS are an ideal setting to investigate magmatic processes and metal transfer mechanisms in a mineralizing magmatichydrothermal system.Combining in situ magnetite trace element contents with whole rock data and modelling provides information on metal behavior in the magma during differentiation.Our data indicate that early and long-lasting sulfide saturation in the magma at Kolumbo leads to the partitioning of chalcophile elements into sulfide phases with limited settling.We suggest that most of the sulfides remain in the magma, either because of their small size (< 10 µm) and high viscosity of the magma, or because of sulfide-volatile compounds formation.Upon degassing, As, Ag, Au, Cu, Hg, Sb, Sn, Pb, and Zn are depleted in the magma, likely partitioning into the volatile phase, either from the melt or during sulfide oxidation by volatiles.After degassing, the remaining chalcophile elements in the melt are incorporated into magnetite as the melt becomes sulfur undersaturated.The metals lost during magmatic degassing are enriched in the mineralization, suggesting a magmatic origin.Chimneys at Kolumbo are also enriched in Tl, which is not depleted in the magma during degassing, and reflect that other metal sources and processes, such as leaching of country rocks, may contribute to the metal budget.Our study shows that trace element contents in igneous magnetite are a powerful tool in tracking metal behavior during magmatic differentiation, identifying sulfide saturation in the melt, and discriminating between pre-and post-degassing magnetite.
Supplementary information.

Fig. 1
Fig. 1 Geological setting of the Kolumbo volcano.a Location in the Anhydros basin-CSK rift: Christiana-Santorini-Kolumbo rift; KVC: Kolumbo volcanic chain; inset: location of Kolumbo within the Hellenic Volcanic Arc.b Bathymetric map (modified after Nomikou et al. 2022) showing sample locations (samples from Klaver et al. 2016 in gray).c Schematic cross-section of the magmatic-hydrothermal system (see b for location), stratigraphy of the K1 to K5 volcaniclastic units based on Hübscher et al. (2015)

Fig. 2
Fig. 2 Petrography of volcanic rocks.a Basaltic andesite lava sample with an amphibole xenocryst.b Trachyte pumice with mafic enclave.c,d Rhyolite pumice with banded texture and mafic enclaves with chilled margin.e Mafic enclave with disseminated magnetite and sulfides in andesite matrix with disseminated magnetite.f Idiomorphic magnetite with sulfide inclusions.g, h Phenocryst clusters with plagioclase, amphibole and magnetite, locally containing sulfide inclusions.i Sulfide bleb of pyrrhotite and chalcopyrite with oxidized goethite rim in basaltic andesite pumice.j Pyrite with magnetite rim in rhyolite pumice with chalcopyrite and magnetite.k Disseminated chalcopyrite and magnetite in andesite matrix.Abbreviations: Ap, apatite; Ccp, chalcopyrite; Gth, goethite; Mag, magnetite; Po, pyrrhotite; Py, pyrite

Fig
Fig.3Geochemical melt evolution modelled using COMAGMAT 3.75(Ariskin and Barmina 2004) for different H 2 O content and fO 2 values compared to whole rock data

Fig. 5 a
Fig. 5 a Classification into magmatic and hydrothermal magnetite based on Dare et al. (2014); b V/Sc in magnetite compared to the MgO content of the whole rock.c Modelling of the sulfur concentration at sulfide saturation (SCSS) versus MgO content in the melt

Fig. 8
Fig. 8 Evolution of Cu/Ag compared to MgO content in volcanic rock samples of different arc volcanoes.Niuatahi data are from Park et al. (2015) and Wang et al. (2019), Brothers volcano, Brothers northeast ridge (BNER), and Brothers west ridge (BWR) data are from Brandl et al. (2022), and Nisyros data are from Georgatou et al. (2022).According to Brandl et al. (2022), basalt from the BWR and basaltic andesite from BNER are considered as parental melt analogues for the Bothers dacite

Fig. 9
Fig.9Cu content in igneous magnetite compared to its V/Sc value.Water saturation is given by COMAGMAT 3.75 modelling and placed on the V/Sc axis after calibration with the composition of magnetite's host-rock 1) Metal zoning is common in VMS deposits and is controlled by temperature-dependent sulfide solubility.Copper requires temperatures above 300 °C for significant transport in typical VMS hydrothermal fluid, whereas Pb and Zn remain soluble at lower temperatures of ~ 180 °C (e.g., Hellyer, Rosebery, Mount Chalmer, and Scuddles VMS deposits, Australia; Large 1992).The venting hydrothermal fluids at Kolumbo are as high as 265 °C (Nomikou et al. 2022).Thus, the Zn-Pb(-Au) sulfidesulfate chimneys represent the cooler part of the system, and Cu is likely trapped at depth as the temperature drops below 300 °C.2) Boiling drastically modifies the chemistry of a hydrothermal solution (e.g., pH increase, changes in fO 2 , fS 2 ), triggering oxide, native metals, sulfide, carbonate, and sulfate precipitation (Drummond and Ohmoto 1985; Heinrich 2007).Scales forming in boiling geothermal systems have trace element enrichments consistent with hydrothermal chimneys and reveal that Co, Cu, and Sn are deposited in the highest-temperature scales (290-345 °C), As and Ni in high-to intermediate-temperature scales, and Ag, Pb, Sb, and Zn in lower-temperature scales (220-260 °C) (Grant et al. 2020).Upon boiling, Au precipitates efficiently but is still enriched in high-to low-temperature scales as it is easily remobilized by hydrothermal fluids over a wide range of temperatures (Grant et al. 2020; Williams-Jones et al. 2009).Similarly, volatile elements (e.g., As, Sb, Se, and Te) are easily remobilized, concentrated during boilinginduced element fractionation and eventually enriched in SMS (e.g., Niua South, Tonga; Falkenberg et al. 2021).At Kolumbo, magmatic fluid mixes with seawater-derived hydrothermal fluids at a depth of 1160 mbsl (i.e., 650 m below the seafloor) and a temperature of 270 °C (Rizzo

Fig. 10
Fig. 10 Average metal content in Kolumbo's sulfide-sulfate chimneys and volcanic samples, normalized to trachyte.The trachyte composition corresponds to the onset of water saturation and degassing.Please note that the vertical scale is increased below 1. Nickel and Co are not shown here because of their siderophile nature and strong partitioning in magnetite, causing depletion between trachyte and rhyolite that is not related to degassing

Table 1
Major element compositions of samples from Kolumbo determined by XRF analysis (see ESM 1 for detailed data, location coordinates for each sample, and data for standards used) Data reduction was done using the Iolite software 3DRS plugin.Accuracy and precision for reference materials (< 15% for most elements) as well as limits of detections are detailed in the Supplementary data (see ESM 2).

Table 2
Trace element composition of samples from Kolumbo determined by LA-ICP-MS analysis, except where indicated (see ESM 1 for detailed data, location coordinates for each sample, *Spectrophotometry data; **ultra-low detection of Au LA-ICP-MS data (see Methods).Abbreviation: n.d.,

Table 3
Major element compositions of magnetite (see ESM 2 for detailed data).Note that the data are sorted by increasing MgO content in the

Table 4
Trace elements composition of magnetite (see ESM 2 for detailed data).Note that the data are sorted by increasing MgO content in the host-rock