The Seven Sisters Hydrothermal System: First Record of Shallow Hybrid Mineralization Hosted in Mafic Volcaniclasts on the Arctic Mid-Ocean Ridge

: We document the discovery of an active, shallow, seafloor hydrothermal system (known as the Seven Sisters Vent Field) hosted in mafic volcaniclasts at a mid-ocean ridge setting. The vent field is located at the southern part of the Arctic mid-ocean ridge where it lies on top of a flat-topped volcano at ~130 m depth. Up to 200 °C phase-separating fluids vent from summit depressions in the volcano, and from pinnacle-like edifices on top of large hydrothermal mounds. The hydrothermal mineralization at Seven Sisters manifests as a replacement of mafic volcaniclasts, as direct intraclast precipitation from the hydrothermal fluid, and as elemental sulfur deposition within orifices. Barite is ubiquitous, and is sequentially replaced by pyrite, which is the first sulfide to form, followed by Zn-Cu-Pb-Ag bearing sulfides, sulfosalts, and silica. The mineralized rocks at Seven Sisters contain highly anomalous concentrations of ‘epithermal suite’ elements such as Tl, As, Sb and Hg, with secondary alteration assemblages including silica and dickite. Vent fluids have a pH of ~5 and are Ba and metal depleted. Relatively high dissolved Si (~7.6 mmol/L Si) combined with low (0.2–0.4) Fe/Mn suggest high-temperature reactions at ~150 bar. A  13 C value of −5.4‰ in CO 2 dominated fluids denotes magmatic degassing from a relatively undegassed reservoir. Furthermore, low CH 4 and H 2 (<0.026 mmol/kg and <0.009 mmol/kg, respectively) and 3 He/ 4 He of ~8.3 R/Ra corr support a MORB-like, sediment-free fluid signature from an upper mantle source. Sulfide and secondary alteration mineralogy, fluid and gas chemistry, as well as  34 S and 87 Sr/ 86 Sr values in barite and pyrite indicate that mineralization at Seven Sisters is sustained by the input of magmatic fluids with minimal seawater contribution. 226 Ra/Ba radiometric dating of the barite suggests that this hydrothermal system has been active for at least 4670 ± 60 yr.


Introduction
Modern seafloor hydrothermal systems are a source of Fe-Cu-Zn-Pb-(Au-Ag) and, like their ancient volcanogenic massive sulfide (VMS) analogues, are formed by the discharge of modified seawater-derived hydrothermal fluids heated by underlying magmas that precipitate metal-rich sulfides on, or near, the seafloor (e.g., [1][2][3]). Metals leached from source rocks exert control on the fluid and sulfide chemistry in most hydrothermal systems, with Cu + Fe + Co + Ni being predominant in mafic-ultramafic settings and Zn + Pb ± Ag ± As ± Sb ± Hg ± Ba dominating more evolved volcanic settings in arcs and backarcs [4]. However, leaching of the host rock volcanics alone fails to explain metal contents in some giant or high-grade ancient VMS deposits [2]. Thus, it has been suggested that magmatic fluids may be a critical source of ore metals for seafloor hydrothermal systems and their ancient analogues, as even a small amount of metal-rich magmatic fluid introduced in a seafloor hydrothermal system can play an important role in mineralization [2,5,6]. Magmatic degassing of volatiles and metals occur at, or just below, the seafloor during submarine volcanic eruptions [7][8][9][10][11][12]. More recently, studies have provided evidence for metal transport in the pre-erupted magma and degassing processes beneath seafloor hydrothermal systems [2,[13][14][15][16]. These magmatic signatures have been especially recognized in arc-related submarine hydrothermal systems, some [17][18][19][20] with submarine epithermal style mineralization comparable to subaerial magmatic-hydrothermal systems [4,16,[21][22][23][24][25][26].
Here, we describe the Seven Sisters vent field, which is a shallow (130 mbsl), relatively hightemperature (~200 °C at 130 mbsl) seafloor hydrothermal system hosted by mafic volcaniclasts within a slow spreading mid-ocean ridge on the Kolbeinsey segment of the Arctic Mid-Ocean Ridge (AMOR). At Seven Sisters, we document a spatial and temporal connection between explosive mafic volcanism, magmatic degassing, and creation of magmatic-hydrothermal mineralization with an epithermal-suite signature. These findings may have important implications for ancient VMS deposit classification and exploration guidelines.

Geological Setting
The AMOR, which extends from north of the Arctic Circle at 66° N to the Siberian Shelf in the Laptev Sea, is composed of several shorter ridge segments that, from south to north, include the following: the Kolbeinsey Ridge, Mohns Ridge, Knipovich Ridge, and the Gakkel Ridge [27,28] ( Figure 1A). The AMOR is spreading at ultraslow rates (<20 mm/yr) with spreading rates decreasing northwards from 18 mm/yr at the Kolbeinsy Ridge to 11 mm/yr at the Gakkel Ridge [29]. The southern part of the ridge system is thought to be strongly influenced by the Iceland hot spot, which results in progressive shallowing of the ridge southward.
Over the last two decades, several hydrothermal sites have been found along the AMOR (e.g., [28,30], and references therein) that are hosted by mafic to serpentinized ultramafic lithologies, at various depths with varying local sediment influence.
The Kolbeinsey Ridge is a ~540 km long super-segment that extends from the Tjörnes Fracture Zone north of Iceland (~66.5° N) towards the Jan Mayen Fracture Zone at 71° N ( Figure 1A). This southernmost part of the Arctic MOR system is characterized by an anomalously thick crust (7-10 km) [31] and a neovolcanic zone that is unusually shallow (<1000 mbsl). The super-segment is itself divided into three segments (i.e., south, central, and north) that are separated by two non-transform offsets at 69° N and 70.7° N, respectively [32].
The Seven Sisters vent field is located on the Northern Kolbeinsey Ridge within an especially shallow part known as the Eggvin Bank. A recent ocean bottom seismometer (OBS) survey across the Eggvin Bank documented an igneous crustal thickness of up to 13 km [33]. The shallowest part of the Eggvin Bank is less than 30 m deep and is defined by a large volcanic edifice with a summit caldera [34]. Large, recent lava flows fill the rift valley to the north of the summit caldera. The Seven Sisters vent field is located approximately 20 km north of the summit crater at a depth of 130 m along a chain of flat-topped central-type volcanic edifices that follow the eastern margin of the rift valley ( Figure 1B,C).

Materials and Methods
Seven Sisters hydrothermal system was discovered in 2013 by a team of scientists from the University of Bergen, with the field revisited and further sampled in 2014 [35]. During both cruises, the R/V G.O. Sars mapped the hydrothermal field with multibeam sonar, and further explored the area with both an autonomous underwater vehicle (AUV) and a remotely operated vehicle (ROV).

Mapping
Bathymetric soundings of the Seven Sisters vent field were collected using the EM302 multibeam echo sounder (Kongsberg Maritime AS, Kongsberg, Norway) onboard the R/V G.O. Sars. All data were acquired using Kongsberg Seafloor Information Systems (SIS) software and included backscatter reflectivity and water column soundings. The data were then processed in Caris HIPS (Hydrographic Information Processing System) software suite using a standard hydrographic workflow. The water column data were processed and visualized in the Fledermaus FMMidwater package, typically in a 'stacked' view that collapsed all water column soundings to an integrated view at the nadir position of the line. This allowed the relatively disorganized bubble plumes found at Seven Sisters to be readily identified, and then more accurately located to identify the individual vent sites prior to AUV and ROV exploration. To reveal and characterize seabed fluid flow features, we collected sonar data using the HiSAS 1030 (2013) and HiSAS 1032 (2014) interferometric synthetic aperture sonar (SAS) [36] carried by the FFI Hugin 1000 AUV. The HiSAS system is a dual-stave interferometric sidescan sonar that operates at a centre frequency of 100 kHz and a bandwidth of 30 kHz. During data acquisition, the Hugin was pre-programmed to travel in a survey grid pattern at a speed of approximately 1 m/s, following the terrain and mapping the seafloor systematically from an altitude of 20-30 m above the seafloor. A high-resolution SAS image is formed by combining dynamically focused data from multiple along-track pings. This creates a sonar image of the seafloor with a constant resolution, independent of the sounding distance from the vehicle. The SAS imagery was produced using the FOCUS processing toolbox [37], and the Kongsberg Maritime-developed post processing software Reflection was used to combine multiple SAS images into geo-referenced mosaics of large areas of the seafloor.

Rock and Fluid Sampling
Dredging operations (referenced as 13DR1, GS14DR2, and GS14DR3) each sampled >50 kg of rocks from the western, northern, and eastern flanks of the field, respectively ( Figure 1C; Table 1). Thirteen rock samples were collected from the active part of the Seven Sisters vent field using the ROV (Table 1). ROV dive GS14ROV02 surveyed the GS14DR2 dredge area; ROV dives GS14ROV03 and GS14ROV14 to GS14ROV18 surveyed and sampled rocks, fluids, and gases from the smoking craters, as well as mounds and venting pinnacles on the summit of the Seven Sisters flat-top volcano. Sample GS14ROV14R1 was collected from the top of Hodr (Höðr) pinnacle. Samples GS14ROV14R2/R3/R4 were taken from the Baldr (Balðr) pinnacle's top, middle, and base, respectively. Samples GS14ROV15R1 to R4 were taken from a phase-separating venting mound named Lily mound, together with samples GS14ROV15R5 and R6, the latter being more distal to the active venting orifice.

Geochemical and Mineralogical Analyses
All rock samples were carefully washed with fresh water. Fresh volcanic whole-rock samples were cut using a rock saw and the least altered portions were selected for geochemical analysis. Mineralized rocks were dried in an oven (max. temp. 48 °C) and stored in sealed bags with nitrogen to prevent further oxidation. Fifty-four polished thin sections were prepared and studied under transmitted and reflected light petrographic microscopes. Representative polished sections were carbon-coated and imaged using the University of Bergen's scanning electron microscope (SEM) Zeiss 55 VP Field Emission, complemented by Thermo Noran Liquid N2 Free detectors allowing for X-ray microanalysis by means of solid-state energy dispersive spectrometry (EDS) (University of Bergen, Norway). Extensive, semi-quantitative SEM-EDS in situ analyses were made of fresh glass shards, sulfides, sulfates, and amorphous silica using Thermo Noran System VI to identify mineral phases too small (<10 m) for unequivocal detection using simple petrographic microscopes.
Whole-rock geochemistry analyses were undertaken by Actlabs Ltd., Canada (www.actlabs.com) using sodium peroxide fusion followed by induced coupled plasma optical emission spectrometry (ICP-OES); aqua regia digestion followed by ICP-OES, ICP-mass spectrometry (MS), and INAA; and cold vapor FIMS. The procedure followed for rock pulverization was as follows: (1) selected samples were cut by diamond saw into small slices <1.5 cm thick, (2) samples were carefully washed and dried at 25 °C, (3) dried samples were crushed using a hammer while enclosed within thick paper envelopes to prevent metal contamination, and (4) small fragments were then powdered in an agate ring mill. Selected samples were manually pulverized using an agate mill and pressed into tablets for analysis using a Bruker D8 X-ray diffractometer at the Department of Chemistry, University of Bergen. Raw spectra and mineral identification were processed in-house using International Centre for Diffraction data (ICDD) mineral database software and later using a different phase identification software, Match! {MatchPhaseidentify}, giving similar results, which are presented as Supplementary Material (File S1).

Isotope Geochemistry
Twelve Sr isotope analyses (including two duplicates) of barite-dominated bulk-rock samples were done at the Department of Earth Sciences, University of Bergen. All samples were heated to 1000 °C for one hour in a Carbolite chamber furnace to remove organic impurities. Samples weighing 100 mg each were digested in concentrated HF on a heating plate at 135 °C for 48 h. The HF supernatant was evaporated to dryness and the residues were then hydrolyzed in a weak solution of HNO3 on a heating plate under sub-boiling-point conditions. This nitrate salt residue was then also evaporated to dryness and completely dissolved in 1 mL 3N HNO3. Strontium was separated by specific extraction chromatography using Sr-spec resin and analyzed using a Finnigan MAT 262 thermal ionization mass spectrometer. Strontium isotopic ratios were corrected for mass fractionation using a 88 Sr/ 86 Sr value of 8.375209. Repeated measurements of strontium carbonate standard SRM 987 during the analysis yielded an average 87 Sr/ 86 Sr value of 0.710235 ± 9 (2σ; n = 10).
Residual barite (BaSO4) from the digestions used for Sr isotope analysis was separated from the solutions by centrifugation in 2 mL micro-centrifuge tubes and sent to ACTLABS for  34 S analysis. In addition, in situ sulfur isotope analyses ( 32 S, 33 S and 34 S) were done on 15 pyrite grains from sample 13DR1R1 using a CAMECA IMS-1280 at the Swedish Museum of Natural History (Nordsim facility), Stockholm, Sweden. Sulfur isotopic analyses were performed on polished rock chips mounted in 25 mm diameter epoxy blocks with a 30 nm gold coating. A section was cut from the sample block to allow co-mounting with epoxy-embedded pyrite standards. Measurements followed the analytical protocol of [38] with a 10 kV and ~2 nA Cs + focused primary beam yielding an average spot size of 10 µm. Gold coating was removed during a 90 s pre-sputter over a 20 × 20 µm rastered area, and detector backgrounds were measured for 30 s. Charge build-up on the sample was prevented using a low-energy electron gun. Instrumental mass fractionation was determined by bracketing analyses of unknowns with measurements of Ruttan pyrite, using a δ 34 S value of 1.408‰ ± 0.006‰ [39]. Sulfur isotopic compositions were calculated as δ 34 S = [(( 34 S/ 32 S) sample/( 34 S/ 32 S) V-CDT) -1] × 1000. Analytical precision and data quality are the same as those reported for samples in [40], which were run during the same analytical session.

Geochronology
Radium isotopes 226 Ra and 228 Ra (with half-lives of 1600 and 5.75 yr) are formed by radioactive decay of 238 U (via 230 Th) and 232 Th, respectively. Because Ra and Ba are chemically similar, they coprecipitate as barite when hot hydrothermal fluid discharges into and/or mixes with ambient seawater. From the onset of mineralization, 226 Ra and 228 Ra decay because they are no longer supported by their radiogenic parent isotopes, 230 Th and 232 Th, as these are not extracted from rock by hydrothermal interaction [41]. Barite is resistant to seafloor (chemical) weathering so that the decaying 226 Ra and 228 Ra are undisturbed from the time of barite precipitation, making 226 Ra dating of hydrothermal barite an important tool in the investigation of temporal evolution of seafloor hydrothermal systems [16,20,42,43]. In this study, ages for barite mineralization in nine representative samples were determined using the short-lived isotope systems 228 Th/ 228 Ra and 228 Ra/ 226 Ra following the method of [42]. The activity ratios of 228 Th/ 228 Ra, 228 Ra/ 226 Ra (Bq•Bq -1 ) and 226 Ra/Ba (Bq•g −1 ) are used for dating barite in the age ranges 0.3-12, 3-35 and 500-15,000 yr, respectively. The initial 226 Ra/Ba value at any one site, that is, the value at the onset of mineralization, must first be established as it is essential for calculating the age of older mineralization using the decrease in 226 Ra/Ba owing to radioactive decay. Recent studies have shown that the barite present in hydrothermal deposits can have a complex history and that they may form during two or more hydrothermal events with different initial 226 Ra/Ba values. In particular, dissolution and precipitation (remobilization) of barite within the deposit is problematic as barite that gain some or all of their Ba and Ra through such a process may obtain significantly lower initial 226 Ra/Ba values owing to decay of 226 Ra in the "old barite" that is being remobilized [44]. Ideally, if there are sufficient samples that still contain 228 Ra, an age for the remobilized older barite can be estimated by plotting a 226 Ra/Ba versus 228 Ra/Ba mixing line for two end members that are different in age [44].

Fluid Chemistry and Dissolved Gasses
Measurements of pH, alkalinity, and various compounds (H2S, NH4, NO3, NO2 and PO4) were made shipboard immediately after fluid sampling. The pH was measured using a portable pH-meter (Metrohm 826) (Metrohm, Herisau, Switzerland), alkalinity was determined by titration using an autotitrator (Metrohm Titrando 888), and the compounds were analyzed by spectrophotometric methods using a QuAtro continuous flow analyzer (SEAL Analytical). Filtered (≤0.2 µm) samples were split into aliquots for later analyses of major (Na, K, Mg and Ca), minor (Si, Mn, Fe and Sr), and trace elements (Ni, Co, Cu, Cd, Zn and Pb) that were filled in acid-cleaned HDPE bottles, acidified to 3 vol% HNO3 and stored at 4 °C until analysed by induced coupled plasma optical emission spectrometry (ICP-OES) (Thermo Finnigan Iris) (Thermo Fisher Scientific, Waltham, MA, USA) and high resolution ICP mass spectrometry (HR-ICP-MS) (Thermo Finnigan Element 2). Aliquots for anions (Cl, SO4 and Br) were stored at 4 °C in HDPE bottles until analysed by ion chromatography (Metrohm).
Dissolved gases and free gas bubbles were collected using titanium alloy gas-tight bottles connected to a snorkel, or funnel inlet, operated by the ROV. After each dive, the gas-tight samples were processed on a high vacuum line to extract the gas phase and subsample for gas analysis later onshore [45]. Compositional analysis of the gas samples was conducted using gas chromatography at the University of Washington, Seattle. Standard errors for CO2, CH4, H2, N2, O2, Ne and Ar were ±3%-5% of the measured value. Extrapolation of the data through zero Mg removes the effects of seawater contamination and provides estimates of the end-member fluid composition [46].
The isotopic compositions of carbon species were determined using a gas chromatograph interfaced with a Thermo ConFlo IV to a Thermo Delta V Plus isotope ratio mass spectrometer at the Geological Institute, ETH Zürich, Switzerland. Isotope values for CH4 and CO2 are reported with respect to the standard Vienna Pee Dee Belemnite (VPDB) in the familiar δ notation. Standard deviations are 0.1‰ for δ 13 CCH4 and 0.2‰ for δ 13 CCO2. Helium isotopes were determined using a 21 cm radius, dual-collector, sector-type instrument specially designed for helium isotope analysis. The measurements were standardized using a Pacific marine air standard. The 3 He/ 4 He ratios are given in R/Ra corr (R = 3 He/ 4 He and Ra = R air = 1.4 × 10 −6 ), which is the value corrected for atmospheric helium contamination. Precision for 3 He/ 4 He values averaged about 0.05 R/Ra. All methods used to determine the gas composition are described in more detail in [47].

Vent Field Architecture
Diffuse venting is widespread on top of the flat top volcano that hosts the Seven Sisters hydrothermal system, although there are three geomorphologically distinct areas that host noticeable hydrothermal activity ( Figure 1C): (1) a summit depression (Figure 2A Figure 2E,F). The summit depression is a collapsed structure exposing fractured crusts of volcaniclasts. Gas bubbles and diffuse venting with associated bacterial mat seeps preferentially follow(?) N-S fractures in layered, coarse-grained consolidated volcaniclasts (e.g., Figure 2B). Ochre and white substrates indicate pervasive hydrothermal alteration, weathering, and leaching. Small mound structures, approximately 5 m in diameter and oriented NE along an alignment that begins at the summit depression, vent phase-separating clear fluids at 199 C and 130 m deep (e.g., Lily mound, Video S1). Some of the larger mounds are partially collapsed, possibly owing to anhydrite dissolution. Outside the hydrothermal mound areas, dark volcaniclastic material fills fractures suggest recent explosive volcanic activity. The northern flank of the flat top volcano reveals a gravitational collapse (not shown) that exposes altered units with unconformable crusts of layered volcaniclasts on top.

Basaltoid Scoria
A vitrophyric, highly vesicular fresh basalt scoria fragment was sampled on the eastern flank of the volcano. Plagioclase phenocrysts are dominant, ranging from a few glomerocrysts up to 1.5 mm in width, to euhedral micro-laths measuring less than 100 m. Subhedral to euhedral plagioclase, and subhedral olivine and clinopyroxene phenocrysts (~500 m width) with corroded rims, suggest meltcrystal disequilibrium. The groundmass is glassy and host vesicles with a bimodal size distribution with approximate diameters of 500 m and 100 m, suggestive of a two-stage vesiculation event.

Consolidated Volcaniclasts with Vitriclastic Textures
A relatively fresh, laminated (<1 cm thick), clast-supported, shard-rich tuff was sampled on the northern flank of the volcano ( Figure 3A). The well-sorted layers are of uneven thickness and alternate between darker, finer shards and lighter, coarser-grain shards. The shards are mostly cuspate ( Figure 4A), but also platy and scoria-shaped, some with contorted vesicles, and may host minor phenocrysts (i.e., ~5% plagioclase; <1% olivine and pyroxene) or primary magmatic sulfide droplets. The glass is isotropic and yellow in thin section, typical of sideromelane. Mn-crust covers the consolidated volcaniclasts and contains two unidentified Mn-oxide phases forming fine, botryoidal, concentric aggregates ( Figure 4B). These glass shards are similar to those found in the mineralized breccias described below.

Mineralized Samples
Mineralized polymictic scoriaceous breccia, vuggy barite-pyrite, and vuggy silica-pyrite materials make up both pinnacles and mounds ( Figure 3B-H) and are ubiquitous in dredged samples. These rocks are matrix-supported and polymictic, depicting varying degrees of hydrothermal alteration and seafloor weathering (e.g., Figure 3F,G).
The mineralization sequence is similar in all the studied mineralized lithotypes. That is, varying degrees of sulfide-sulfate mineralization are seen, ranging from less than 5% to almost complete replacement of the volcaniclastic rock. Figure 3H shows a mineralized rock in which a sulfide mineralization "front" (bottom grey) moves upwards, replacing altered volcaniclastic material (pale green). The top of the sample is composed of weathered material exposed to the seafloor.
Barite and minor anhydrite are the major hydrothermal mineral phases present in the samples. Anhydrite crystals are euhedral, presenting as large prisms ( Figure 4C,D), whereas barite crystals are euhedral, but show different growth habits from early, large, tabular crystals ( Figure 4E), to sheathlike to late-stage radial fibrous arrangements ( Figure 4F).
Pyrite is ubiquitous, preferentially found in the vesicles of the glass shards ( Figure 4G,H) and rimming grain boundaries ( Figure 5A), and is the first sulfide to form ( Figure 5B). Pyrite can be euhedral, pseudomorphic lamellar after barite, and framboidal. Small (10 m), cubic pyrite crystals grow inside the volcaniclastic vesicles until the original vesicular texture is entirely replaced by massive pyrite up to 2 mm in size ( Figure 4G,H). Larger pyrite grains are zoned, subhedral, and intergrown with marcasite. A second stage of sulfide mineralization is characterized by the occurrence of sphalerite, chalcopyrite, galena, and Ag-sulfosalts ( Figure 5C-I). These phases are typically smaller, between 10 and 20 m, and mantle pyrite rims ( Figure 5C,D) precipitate after barite ( Figure 5E,G) or are characterized by fine, dendrite-like disseminations within spherules of micro and cryptocrystalline silica ( Figure 5H,J). Fe-poor, translucent sphalerite grains, commonly associated with chalcopyrite, are anhedral when rimming pyrite and barite. However, if sphalerite has grown into open spaces, it tends to be spherical in habit, exhibiting botryoidal growth ( Figure  5H).  Subhedral galena appears as epitactic growths on pyrite grains ( Figure 5C,D) or intergrown with chalcopyrite ( Figure 5F) and is associated with Au and Ag peaks on EDS spectra. Other Au peaks detected in the EDS spectra were associated with amorphous silica, barite, and chalcopyrite. Sulfosalts have been identified through SEM-EDS semi-quantitative analysis as argento-tennantite and argyrodite ( Figure 5I). In addition, there are other very small (less than 1 m wide) unidentified sulfide phases bearing Pb-Ag, As-Pb, Bi-S, Tl, Mo and Au.

Silica Crusts
A dredge on the northern flank of the flat topped volcano recovered a composite sample that included a sacharoid-vuggy, silica-rich layer lying between coarse-grained layers of volcaniclastic material and an Mn-like crust. Smaller in size, but ubiquitous in dredge GS14DR02, were white and grey deposits of silica ( Figure 3I, sample GS14DR2R1 and GS14DR2R2). This may correspond to some of the crusts seen around the active vent areas from which warm and diffuse fluids were seen to be discharging. These sinter-like samples are mostly composed of sacharoid, colorless silica interlaid with sub-metallic grey silica ( Figure 3I). Silica micro-textures are characterized by both straight and curved 10 m wide filamentous silica networks of radiating rods, and small (up to 10 m) spherules of amorphous and cryptocrystalline silica ( Figure 5J).

Hydrothermally Altered Material
Dredge 13DR01 sampled many granular, earthy and friable, pale yellow ( Figure 3J) to ochre clay-rich samples. These textural features are similar to some of the coarser grain breccias (e.g., Figure  3F), suggesting that some mineralized breccias underwent intense leaching that obliterated most of the primary volcanic-sedimentary textures.

Geochemistry of Mineralized and Hydrothermally Altered Samples
Whole-rock compositions of representative mineralized samples ( Table 2) show that most are low in sulfide sulfur and base metals (i.e., Cu, Zn and Pb) and that silica and Ba are major components with contents of up to 60.5 wt.% and 43.8 wt.%, respectively. Other, typically common major elements such as CaO, MgO, Na2O, and K2O are only minor in these samples.
Iron correlates positively with S and is most abundant in the sulfide-rich samples (up to 35.1 wt.% FeO; 33 wt.% S) owing to the pyrite-dominated mineralogy. Pyrite-rich samples are Badepleted, but show higher As, Hg, Mo, Tl and Sb contents. Layered sample GS14ROV14R4, subdivided in -b for the silica-rich layer and -y for the pyrite-rich layer, contains the highest As With the exception of high SiO2 contents (70.6 wt.%), silica crust material is depleted in other major oxide elements. Barium is the second most abundant element (6.29 wt.%) and yet still represents <10% of most silica crusts. Barite grows as euhedral radial blades (rosettes) dispersed within the silica, or as laths aligned in layers. Notwithstanding its low sulfide S content (0.12 wt.% S), the Si-crust sample contains 15 ppm Ag and base-metal contents similar to those found in samples with visible sulfides. The dominant grey layers seen in dredge sample GS14DR2R2 (e.g., Figure 3i) show the presence of minute sulfides coating the silica rods, visible under the petrographic microscope.
X-ray diffraction analysis on bulk powder (File S1) confirmed the mineralogical observations and, furthermore, identified clay minerals (kaolin-group minerals, namely dickite and nacrite) and Si-phases (possibly cristobalite and other SiO2 phases) not easily classified using standard petrographic techniques. Barite, including a Pb-rich barite variety (also confirmed using SEM-EDS), is ubiquitous in the analyses, followed by pyrite and marcasite as the main sulfide phases. Anhydrite was identified in sample 13DR1R5 ( Figure 3H), consistent with petrographic observations.

Isotope Geochemistry
The results of whole-rock strontium isotope ratios ( 87 Sr/ 86 Sr) of barite-rich samples and bulk sulfur isotope analyses of barite separates from mineralized samples and silica crusts are given in Table 3. Values for 87 Sr/ 86 Sr show a small range between samples (Figure 6a). from 0.703961 ± 8 to 0.704834 ± 8 (average 0.704308) and lie between modern seawater values and Northern Kolbeinsey Ridge basalt at ~71 N ( 87 Sr/ 86 Sr = 0.703047; [48]). Sulfide-rich mineralized breccias, which contain native sulfur in cavities, are the most radiogenic samples ( 87 Sr/ 86 Sr = 0.704834), whereas the least radiogenic material is the silica crust with 87   Barite  34 S values range between 17.8‰ and 24.0‰, ranging to both more negative and positive values relative to modern seawater sulfate (21.5‰) [49]. Lower than seawater  34 S values are found at the base of Baldr pinnacle, in silica crust, and the Lily mound. Higher  34 S values are from sulfatebearing samples collected from the top of the Lily mound and from the Hodr and Baldr pinnacle vent conduits.

Geochronology
In total, nine samples were used for radiometric dating, which include a Si crust, the top of Hodr pinnacle (2 samples), Baldr pinnacle (3 samples), and Lily mound (3 samples; see Table 5). Recent mineralization ages of <12 yr old are given by the activity ratio 228 Th/ 228 Ra, that is, the amount of 228 Th grown from deposited 228 Ra. Lily mound samples range from 3.6 ± 0.03 to 4.4 ± 0.05 yr, whereas samples from the top of the Hodr and Baldr pinnacles have 6.5 ± 0.8 and 4.1 ± 0.6 yr, respectively. The oldest 228 Th/ 228 Ra mineralization age is given by a sample taken for a section close to the Baldr pinnacle top at 8.1 ± 2.6 yr.  [43,50]) and show little influence of seawater [51], but are close to the 87 Sr/ 86 Sr value for local basalts [48] from the Northern Kolbeinsey Ridge (NKR);  34 S in barite has a greater range, plotting above (top of venting structures) and below seawater values ( 34 S 21.5; [49]; (B) in situ  34 S vs.  33 S in pyrite shows MORB-like  34 S close to 0.0‰ (0.34‰ ± 0.64‰) and predominantly negative  33 S values that plot distant from known biogenic pyrite (1; [52]) or hydrothermal pyrite fields (2 and 3; [52,53]). The lowest 226 Ra/Ba values were measured in the Si-crust (6 Bq•g −1 ), Baldr pinnacle base (6.5 Bq•g −1 ), and a sample close to Lily mound (29 g −1 ). No 228 Ra was detected in these samples. Samples from the top of the actively venting pinnacles and mound have higher 226 Ra/Ba values (35-49 Bq•g −1 ) with the highest value (49.0 ± 0.5 Bq•g −1 ) measured from the white section of Hodr's pinnacle top ( Figure 3B), corresponding to material least affected by remobilized older barite, and thus decaycorrected to the time of mineralization. This corrected initial 226 Ra/Ba value (49.14 ± 0.51 Bq•g −1 ) is taken as the initial value when applying the dating method of decreasing 226 Ra/Ba values with time. Dividing the 226 Ra/Ba value of a sample by this initial ratio gives the decay factor used to calculate the age of the older components. However, the ages derived using 226 Ra/Ba values appeared to be much older when compared with those using 228 Th/ 228 Ra values (e.g., 350 ± 40 yr vs. 4.4 ± 0.05 yr at Lily mound; GS14ROV15R1). However, by combining the Ba, 228 Ra, and 226 Ra data, we find that these older ages are almost certainly incorrect owing to an inherited component of much older, remobilized barite, and thus an age for the older barite component can be estimated [44].  [57,58].

Fluid and Gas Chemistry
End member CO2 concentrations are 2.94 mmol/kg for the Lily mound (ROV18-GT16) and 5.58 mmol/kg for the Hodr pinnacle (ROV15-GT16). End member CH4 and H2 concentrations were low in both vents and ranged from 15 to 26 µmol/kg and 4 to 9 µmol/kg, respectively. Total measured gas concentrations were 3 mmol/kg in the phase-separated vent and 5.1 mmol/kg (corrected for air) in the Hodr pinnacle. A bubble sample collected near Lily mound (ROV16-GT17), an area where bubbles were seen seeping from the seafloor, was dominated by 80.3 % CO2, and further contained 4% CH4 and 0.01% H2. Carbon isotope values for Hodr vent are δ 13 CCH4 = −11.1‰ and δ 13 CCO2 = −5.4‰. The isotopic values of the bubble sample, δ 13 CCH4 = −8.1‰ and δ 13 CCO2 = −1.7‰, were more positive when compared with the Hodr fluid sample. The 3 He/ 4 He values in the Hodr pinnacle vent and Lily mound were 7.8 R/Racorr and 8.6 R/Racorr, respectively. Similarly, the bubbles near Lily mound had a 3 He/ 4 He value of 8.6 R/Racorr. Prior to end-member calculation (Mg = 0), the lowest Mg concentration found was 1.6 mmol/L in the sample from the phase-separated vent, indicative of a reasonably pure end member sampling.

Explosive Volcanism, Mafic-Volcaniclastic Sucessions, and Hydrothermal Activity
The Seven Sisters hydrothermal system sits on one of the flat-topped, N-S aligned shallow volcanic edifices and likely originates from the same magmatic source as the adjacent summit caldera, known as the Eggvin Bank. These flat-topped volcanoes are probably cogenetic, erupting from a dikelike structure or/and from a single magma lens. The Seven Sister's proximity to Eggvin Bank, combined with extensive volcanism within the surrounding areas, is characteristic of relatively high magma supply [33]. Fresh, mafic, and highly vesicular basaltic scoria sampled from the flat-topped volcano that hosts the Seven Sisters magmatic-hydrothermal system indicates recent pre-to syneruptive degassing of exsolving volatiles in magma that was readily quenched at the seafloor. However, the most common volcanic lithotype is unconsolidated and consolidated volcaniclastic material that is ubiquitously distributed throughout the Seven Sisters site and that shows different grain-size distribution ranging from predominantly fine-ash to the north (e.g., Figure 4a), to coarse, 6 mm lapilli-tuff to the east. These volcaniclasts are evidence of episodic and explosive volcanism with load compaction of hot plastic glass (e.g., [67]), while their polymictic nature indicates some degree of post-depositional reworking. Explosive disruption of vesicles makes glass shards very susceptible to alteration so that the non-altered character of some samples suggests recent explosive volcanism. The laminated appearance of the consolidated deposit is the result of layers of partially oxidized material, indicative of seafloor weathering prior to further episodic explosive events, leading to the deposition of a new layer of glass shards.
At Seven Sisters, hydrothermal activity is pervasive on top of the volcanic edifice, either as diffuse venting with gas seeping through fractures covered by bacterial mats, bubbles seeping through the seafloor or, as focused, up to 200 °C venting from the clusters of craters, mounds, or pinnacle-like structures. The latter differs from a typical vent chimney as they are made up of products of vuggy barite-silica-pyrite and lack a well-defined central venting orifice. Similar structures have been observed at the Giggenbach and Clark volcanoes of the Kermadec Arc [68].
In ancient volcaniclastic-dominated VMS deposits, the diffuse discharge of hydrothermal fluids through permeable volcaniclastics, eventually capped by a silicified horizon, resulted in sulfide mineralization occurring by replacement, infilling pore spaces, and producing tabular subseafloor VMS deposits [69]. Under shallow-water, some of these volcaniclastic-dominated VMS deposits may be characterized by transitional features between VMS and epithermal deposits (e.g., aluminumsilicate or K-bearing alteration assemblages) [69,70]. Similarly, at Seven Sisters, seafloor venting of Ba-and Si-rich hydrothermal fluids in a non-consolidated, reworked, volcaniclastic deposit ensured rapid precipitation of barite (Figure 4f; [71]) and silica when the hydrothermal fluid mixes with seawater with concomitant sealing and insulation of the top layer. At Seven Sisters, the volcaniclasticdominated lithologies are mafic in composition, making this a relatively unique shallow, mafic, volcaniclastic-hosted seafloor hydrothermal system.

Fluid Chemistry and Phase Separation Processes in a Shallow System
Fluid-rock interaction and phase separation can control vent fluid chemistry, resulting in a wide range in end-member compositions. This is particularly true in shallow systems, regardless of the setting, where lower pressures (and temperature) promotes boiling, which in turn exerts a strong control on nature of metals to be transported in solution (e.g., [72]). Thus, metals that are transported at relatively lower temperatures, such as Ag, Au, As, Sb, Hb, Tl, Pb and Zn, tend to precipitate at shallower depths and above boiling zones [63], making shallow hydrothermal systems favorable settings for the precipitation of sulfides and sulfosalts containing these elements.
During phase separation (boiling), most elements are conservative relative to Cl, whereas dissolved gases (e.g., CO2, CH4, He, H2, and H2S) that do not form chloro-complexes segregate preferentially into the low Cl phase, that is, the vapor phase [59,73,74]. Changes in element ratios with respect to Cl provide information on the extent of the water-rock interaction (e.g., [58]). Fluids expelled from the Hodr pinnacle contain lower chlorinity and higher gas contents, suggesting that phase separation occurs beneath the structure. Moreover, low Na/Cl and high Ca/Cl and K/Cl values for these same vent fluids are indicative of Na uptake and Ca and K release reactions subseafloor. The depth (~130 m) and temperature (~200 °C) at which Lily mound fluids were sampled places them on the two-phase boundary for seawater (e.g., [75][76][77]), confirmed by the "flaming" effect caused by phase-separation processes recorded by the ROV cameras (Video S1).
Phase separation is an important mechanism for sulfide precipitation owing to cooling by decompression and an increase in pH and fO2 following the loss of gas [78], which will physically limit venting temperatures [79], thus simultaneously influencing the metal transport capacity in the vapor and fluid phases [80][81][82]. Not surprisingly, the Seven Sisters vent fluids are metal-depleted when compared with a typical 1500-2500 m deep, higher temperature system (e.g., [20,55,77,83]). Nevertheless, fluids with relatively high Zn concentrations (200x that of seawater) vent from the mound. Arsenic found in Seven Sisters fluids ranges from 315 to 493 nmol/l, higher than As found in fluids from the Mid-Atlantic Ridge vent fields (<319 nmol/l) [84]. Vent fluids with higher As values have to our knowledge only been reported from hydrothermal systems in back-arc environments such as the Lau Basin [84].

Silica and Secondary Alteration Minerals
In shallow magmatic-hydrothermal systems, acid leaching alters feldspars producing kaolinite under low temperatures (<120 °C), and dickite at higher temperatures (150-350 °C) [85][86][87][88], making this assemblage, along with the presence of native sulfur, a key indicator of solutions with low pH [89]. Gangue minerals such as kaolinite-group minerals and vuggy silica were found at Seven Sisters, especially in the extensively altered volcaniclastic material. Similar alteration mineralogy has been previously described in seafloor hydrothermal systems associated to arc and back-arc settings (e.g., [16,20,90]).
Cryptocrystalline and microcrystalline silica phases are abundant at Seven Sisters, occurring either as sinter-like Si-crusts (containing Ag-Pb), or as a late-stage phase in volcaniclastic breccias mantling dendritic sulfides and sulfosalts. Late-stage filamentous spherules of cryptocrystalline silica and similar amorphous silica materials have also been described in submarine arc, back-arc, and MOR hydrothermal systems [50,[91][92][93][94][95][96]. Under hydrothermal conditions, silica solubility decreases with decreasing temperature and pH, thus supersaturation and precipitation of silica occur preferentially by cooling prior to mixing with seawater [96][97][98][99][100]. At Seven Sisters, the lowest 87 Sr/ 86 Sr values (i.e., indicative of minimal mixing with seawater) correspond to these sinter-like materials (Table 3; Figure 6), consistent with a relatively pure end-member mineralizing fluid that leads to silica precipitation.
At 130 m water depth, boiling occurs at ~200 °C [75]. Cooling and supersaturation of silica and formation of silica colloids requires an initial fluid that is in equilibrium with quartz deeper in the system, at higher temperatures [94]. Fluids at Seven Sisters show relatively high dissolved Si (7-8 mmol/kg) for venting at ~13 bar and ~200° C. By comparison, silica solubility calculated for seawater at ~150 bar and 200 °C is 4.56 mmol/kg [101]. Estimated minimum subseafloor reaction temperatures using the Fe/Mn geothermometer of [102] give a temperature range between 250 °C for Lily mound and 283 °C for the Hodr pinnacle. Assuming SiO2 saturation at depth, the silica solubility would be ~7.9 mmol/kg at 150 bar and 250 °C [101], a value that is similar to those observed at Seven Sisters. Although considered unlikely, subsurface mixing with a lower temperature fluid in equilibrium with amorphous silica [101] and low water/rock ratios [103] could also contribute to higher Si contents.

Magmatic Degassing at Seven Sisters
Magma degassing is the likely source of CO2 in the Seven Sisters vent fluids. Carbon dioxide is predominant over other gas species found in the volatile fraction ( Table 6). Saturation of dissolved CO2 is about 5 mmol/kg at the depth, temperature, and salinity conditions found at Seven Sisters. Therefore, the formation and venting of CO2-rich bubbles (>80% CO2) near Lily mound suggest CO2 concentrations greater than the solubility at 130 m water depth and a release of additional CO2 to the gas phase.
The Hodr vent fluid δ 13 CCO2 value of −5.4‰ falls in the range of sediment-starved hydrothermal systems, commonly ascribed to a mantle carbon source [104,105]. Furthermore, R/Ra values derived from 3 He/ 4 He measurements for the Seven Sisters vent fluids have MORB-like signatures, again consistent with a mantle source (e.g., [106]). A δ 13 CCO2 value of −1.7‰ for the CO2-rich bubbles venting near Lily mound appears too high to have a mantle-only source and likely reflects additional process(es) contributing to the bubble carbon budget. High δ 13 CCO2 fluids found in hydrothermal systems have been explained by either magma source enrichment due to slab-derived carbonate inputs (e.g., [7,104]) and/or degassing related to fractionation during early stages of CO2 exsolution [107,108]. As degassing progresses, both exsolved CO2 and residual melt become increasingly depleted in 13 C [107,109]. Carbonate input seems unlikely in this setting. On the other hand, magma resupply and degassing appear to be an episodic event within the Seven Sisters volcanic complex. We suggest, therefore, that the δ 13 CCO2 signature in the venting bubbles near Lily mound could indicate direct volatile exsolution from a relatively undegassed magma and that the Hodr fluids tap a mixture between that and reservoir equilibrated volatiles.
Both CH4 and H2 concentrations were low compared with other vent fields along the Mid-Atlantic Ridge, suggesting the hydrothermal system at Seven Sisters has not particularly interacted with sediments [47,110]. This is consistent with in situ pyrite sulfur isotope data (Figure 6), where the δ 34 S values (~0.36‰) cluster around MORB values precluding interaction of bacterially-reduced (i.e., low δ 34 S) sulfur from marine sediments [111].
Conversely, the Seven Sisters δ 13 CCH4 values are relatively high (−11.1‰ to −8.1‰) compared with sediment-starved, basalt-hosted hydrothermal vent systems [112]. The isotopic composition of CH4 precludes a biogenic or thermogenic source for the excess CH4 in the bubbles, and falls in a range typically observed in ultramafic environments originating from abiogenic contributions to the CH4 budget [113], or favoring a process of reduction of mantle CO2 [114,115]. An ultramafic input to the hydrothermal system at Seven Sisters would seem unlikely based on the geological setting. Notwithstanding, high δ 13 CCH4 values at Lucky Strike hydrothermal vent field of the Mid-Atlantic Ridge have been attributed to serpentinization of distal ultramafics [116]. It is more likely that subseafloor CH4 oxidation [117] is the process responsible for the isotopic compositions values at Seven Sisters, which would be consistent with the mineralogical observations described above.

Tracing the Fluid Source at Seven Sisters
Barite is a common gangue mineral in the walls of vent chimneys where extensive mixing of seawater occurs (e.g., [16,94,96,118,119]). At Seven Sisters, ubiquitous barite is the first hydrothermal mineral to precipitate and is the dominant sulfate that builds up the venting structures (pinnacles and mounds), whereas anhydrite is less common and mostly associated with larger grain size volcaniclastic deposits (e.g., Figure 3F). Barite appears dispersed in silica, similarly reported in other shallow submarine hydrothermal vents [120] or as rosettes, and is common within mineralization at Brothers, Clark, and Rumble II West volcanoes of the Kermadec arc, where it is considered indicative of rapid precipitation and/or quenching [20,68,71,119].
Strontium and Ba are efficiently leached from basalts by hydrothermal fluids [103,121], with barite (which contains Sr) precipitating when seawater SO4 2− mixes with Ba 2+ carried by the hydrothermal fluid [94,122]. Consequently, the isotopic signature of 87 Sr/ 86 Sr in barite is a useful tool in identifying the fluid source [7,123,124]. At Kolbeinsey Ridge, seawater has a 87 Sr/ 86 Sr value of 0.709225 [51], while basalts sampled nearby at ~71ºN (i.e., in close proximity to Seven Sisters) are less radiogenic with 87 Sr/ 86 Sr = 0.703047 [48]. The average 87 Sr/ 86 Sr value for Seven Sisters is ~0.704340, close to the isotope signature of local basalt, implying limited interaction with seawater by the hydrothermal fluids (Figure 6a).
The relative contribution of each endmember fluid that leads to barite precipitation can be assessed using a two component mixing model [125]:  [51], barite-rich material (Table 3), and MOR-type hydrothermal fluid [126], respectively. The results support a predominant hydrothermal contribution of Sr (~67% ± 6%) over seawater ranging from 56% in the mineralized breccia to 74% in the silica-crust material.
Hydrogen sulfide produced by bacterial sulfate reduction is characterized by high  33 S and significantly negative and heterogeneous  34 S [53,[128][129][130][131][132]. We suggest that near-zero homogeneous  34 S and slightly negative  33 S in pyrite (Figure 6b) is not compatible with significant biogenic sulfate reduction in the hydrothermal system. Furthermore, oxidation of H2S during phase separation leads to considerable isotopic heterogeneity [133] that is not seen here, suggesting that disproportionation of magmatic SO2 − [134] may be a contributing factor for the  34 S values measured in Seven Sisters. Similar observations were made for the Hina Hine system in Lau Basin [135] and several arc volcano systems. Magmatic SO2 disproportionates to H2SO4, H2S, and S during phase separation (e.g., [16,20]) with co-precipitation of sulfate minerals, sulfides, and native sulfur ensuring significant isotope fractionation between H2SO4 that is enriched in 34 S, and H2S and S0 that are depleted in 34 S (e.g., [134,136]). Moreover, the presence of elemental sulfur infilling vugs (Figure 3e) has been cited as evidence for condensation of magmatic SO2 in association with characteristic magmatichydrothermal mineral assemblages [16,17,137].

Age of Mineralization
Recent barite mineralization true ages were determined using 228 Th/ 228 Ra values and range from 3.6 to 8.1 yr from the time of sampling (Table 5). These ages relate to samples taken from the top of actively venting pinnacles and mound.
The oldest barite mineralization events were estimated using 226 Ra/Ba values with minimum ages of 4840 and 4670 yr attributed to the Si-crust and base of Baldr pinnacle, respectively. Initial 226 Ra/Ba values for barite mirror the initial abundance of 226 Ra and Ba in the source rock; however, ascending hydrothermal fluids that transport the newly extracted Ba (and Ra) may dissolve and remobilize older subseafloor barite (with 226 Ra that has already decayed), lowering the initial 226 Ra/Ba values [42,44,68]. Thus, when dating older barite components, comparing their 226 Ra/Ba values with the maximum 226 Ra/Ba values (i.e., the initial value used for the site) will give minimum ages rather than absolute ages [44].
Barite remobilization concomitant to zone refining processes in mounds and at the subsurface has been described in modern seafloor hydrothermal systems and their ancient VMS analogs (e.g., [43,44,52,68,96]). The observed multi-stage barite growth, if sourced by the remobilization of older barite from the mound and subsurface, may explain the observed wide range of 226 Ra/Ba values and ages at Seven Sisters (Table 5). To validate this assumption, radium isotope activities are decaycorrected, using ages derived from 228 Th/ 228 Ra values, to the time of mineralization for a 226 Ra/Ba versus 228 Ra/Ba mixing line (Figure 7). If barite was produced by separate hydrothermal events, the mixing line will intercept the 226 Ra/Ba axis, that is, give the 226 Ra/Ba value for the older barite component (Figure 7). Its age can then be calculated by comparison with the highest 226 Ra/Ba value, 49.1 ± 0.5 Bq•g −1 , taken as the initial value for mineralization containing minimal remobilized older barite (Table 5). From Figure 7a, the 226 Ra/Ba value for the older barite in the pinnacles is 12.2 Bq•g −1 , which gives an age of 3220 ± 560 yr, and is thus younger than the age estimated for the base of Baldr pinnacle (GS14ROV14R4) of 4670 ± 60 yr. The large error (1σ) relates to the counting statistics on the relatively low 228 Ra activities and the decay corrections to the time of mineralization. The difference in age may be real considering GS14ROV14R4 is 75 m away from GS14ROV14R3 (Baldr Pinnacle), assuming that using data from both pinnacles for the mixing line is valid. The results for Lily mound samples GS14ROV15R1 and GS14ROV15R2 are more reliable owing to smaller errors. From Figure 7b, the 226 Ra/Ba value for the older barite is 30.7 Bq•g −1 , which gives an age of 1080 ± 270 yr, which is comparable to 1200 ± 40 yr for sample GS14ROV15R5, which is close (<70 m) to the mound. The consistent ages derived by these different methods, that is, via the barite mixing line and by comparing the minimum and maximum 226 Ra/Ba values for the mound suite of samples, are evidence that the sample close to Lily mound is genuinely much younger than that of the base of Baldr pinnacle located ~320 m away (Figure 2e). Sample GS14DR2R2, a dredged Si-crust, is similar in age (4840 ± 70) to the base of Baldr pinnacle (Table 5).

A Shallow, Mafic Volcaniclastic Hosted VMS Analogue?
The spatial proximity between some subaerial epithermal and submarine VMS deposits raises the possibility that hybrid VMS deposits with epithermal-style mineralization may occur in shallow waters [70,138]. Shallow submarine hot springs are common in submerged volcanoes on island arcs with known porphyry copper and epithermal gold (e.g., [21,94,[139][140][141]). In submarine settings, however, while high sulfidation and acid-sulfate alteration is common, seawater replaces meteoric water diluting magmatic derived fluids, buffering pH and temperature [70]. One example is the submarine Brothers volcano, where the alteration mineral assemblages suggest a progression from an initially magmatically influenced to a seawater-dominated hydrothermal system [17]. At Brothers, the presence of illite, natroalunite, pyrophyllite, quartz, opal CT, pyrite, and native sulfur indicates high temperature reactions with acid-sulfate fluids. Sinter alteration, common in low-sulfidation deposits and neutral pH hot springs carrying Ag, Hg, Tl, Sb and Au [142], may actually occur in these seawater-buffered systems. The occurrence of these elements has been documented in some VMS deposits and resulted from condensation of ascending magmatic volatiles exsolved during boiling and oxidation of excess S in solution [2,143,144].
Pyrite-rich samples at Seven Sisters are enriched in Tl, Hg, As and Sb and their contents are comparable to those of magmatic-hydrothermal systems like Palinuro (for Tl) and Brothers volcano (for Sb) [16,21] (Figure 8). Few MORB samples also show comparable amounts of Tl (Figure 8) [4], although these samples are considerably more sulfide-rich than their Seven Sisters counterparts, emphasizing the epithermal suite enrichment at Seven Sisters.
Seven Sisters is located in a slow spreading ridge, where volatile-rich basalts and explosive volcanism are genetically related to mineralization, although no secondary carbonate minerals were found. Mineralization temperatures >200 °C are within the epithermal range, whereas the presence of argillic alteration, sulfate-sulfide-sulfosalt assemblages (with Ag, As, Tl, Sb, and Hg), and native sulfur, when combined with S, Sr and δ 13 CCO2 isotopic data in rocks and fluids, is evidence for magmatic input into the hydrothermal system.
Slow to ultra-slow MORB settings like the Northern Kolbeinsey Ridge (NKR) are not expected to host this type of mineralization, although the authors of [138] had hypothesized that even tholeiitic basalt emergent from MORB or hot spots could possess underappreciated epithermal potential. We thus propose that the mineralization style at Seven Sisters displays hybrid characteristics between VMS and seawater-buffered epithermal deposits.

Conclusions
The Seven Sisters is the first known shallow, hybrid, seafloor hydrothermal system with epithermal-style mineralization that is hosted in mafic volcaniclastics on a slow spreading ridge. The northern Kolbeinsey ridge area near Seven Sisters is characterized by enhanced volatile-rich magmatic activity, multiple episodes of explosive volcanism, and formation of volcaniclastic deposits with post-depositional reworking.
At Seven Sisters, hydrothermal fluids percolate through permeable mafic volcaniclastic deposits precipitating sulfate-sulfide minerals by replacement and in-filling of pore spaces. Focused, phaseseparating of clear fluids of ~200 °C seep through unique barite-rich pinnacle-like structures and mounds.
Barite is ubiquitous and replaced by pyrite, which is the first sulfide to form, followed by Zn-Cu-Pb-Ag bearing sulfides, sulfosalts, and silica. Sulfides are scarce, usually less than 10%, resulting in relatively low (<0.1%) base-metal (Cu-Zn-Pb) concentrations. However, whole rock data show high contents of 'epithermal suite' elements such as Hg, Tl, Sb, Ag, Au and Mo. Gangue alteration mineralogy assemblages (e.g., dickite) are suggestive of a magmatic influence on the hydrothermal systems.
End member vent fluids show both lower and higher than seawater chlorinity, have a pH of ~5, and are depleted in metals. Higher dissolved metal contents are found in the mound fluids, where phase separation visibly occurs at the seafloor. Conversely, lower dissolved Cl and H2S, combined with low metal content at the Hodr vent, suggest here that phase separation occurs subseafloor, where sulfides are believed to be precipitating.
Carbon dioxide is expelled as bubbles (>80% CO2) near Lily mound, and CO2 dominates the volatile fraction at Seven Sisters fluids, with  13 C values of −5 per mil as evidence for magmatic degassing from a relatively undegassed reservoir. Low concentrations of CH4 (<0.026 mmol/kg) and H2 (<0.009 mmol/kg) are consistent with a sediment-free hydrothermal system at Seven Sisters, corroborated by 3 He/ 4 He between 7.8 and 8.6 (R/Ra corr.), which shows MORB-like signatures, indicating an upper mantle source.
MORB-like multiple sulfur isotopes in early pyrite indicate little input from reduced seawater sulfate or leached sediments with biogenic sulfide, whereas sulfate  34 S in barite is explained by the precipitation from a fluid containing seawater sulfate mixed with a small portion of lower  34 S derived from the oxidation of magmatic sulfur. 87 Sr/ 86 Sr values in barite-dominated samples are close to the isotope signature of local basalt and indicate little seawater input into the hydrothermal fluids. The least radiogenic sample found at Seven Sisters is an Si crust. This sample has more negative than seawater  34 SSO4 values and is 228 Ra depleted, and may be closer to pristine hydrothermal fluids at the summit of the Seven Sisters.
Hydrothermal barite-rich deposits have been produced, albeit intermittently, at Seven Sisters for at least 4700 yr, which is critical for better understanding the hydrothermal activity and distribution of seafloor mineral deposits in this area.