Effects of Substrate Composition and Subsurface Fluid Pathways on the Geochemistry of Seafloor Hydrothermal Deposits at the Lucky Strike Vent Field, Mid‐Atlantic Ridge

The Lucky Strike vent field, located on the Mid‐Atlantic Ridge (MAR), is hosted on enriched mid‐ocean ridge basalt associated with the nearby Azores hotspot. In this study, we present bulk rock geochemistry coupled with in situ sulfur isotope analysis of hydrothermal samples from Lucky Strike. We assess the geological controls on the differences in the major and trace element content and sulfur isotopic composition of the hydrothermal deposits within the vent field. The hydrothermal deposits contain elevated concentrations of elements typically enriched in mid‐ocean basalt (E‐MORB), such as Mo, Ba, and Sr, compared to typical values for other hydrothermal deposits hosted on the MAR. The range in sulfur isotope compositions of hydrothermal marcasite and chalcopyrite (−2.5 to 8.7‰) is similar to the range recorded at other sediment‐free basalt‐hosted seafloor hydrothermal sites. However, at Lucky Strike, the Capelinhos vent, situated 1.4 km east of the main field, is enriched in 34S (by ∼3.5‰ for both marcasite and chalcopyrite), relative to the main field. This difference reflects contrasting subseafloor fluid/rock interactions at these two sites, including subseafloor sulfide precipitation at the main field that results in <20% of reduced sulfur within the upwelling hydrothermal fluid reaching the seafloor. We also compare the geochemistry of the hydrothermal deposits at Lucky Strike to other hydrothermal sites along the MAR and show that the average hydrothermal deposit Ba/Co is useful to discriminate between E‐MORB and other mafic/ultramafic hosted deposits.

2 of 24 Petersen et al., 2000). However, the deposits at Lucky Strike also contain abundant barite, which Langmuir et al. (1997) link to an enriched mid-ocean basalt (E-MORB) substrate associated with the nearby Azores hotspot.
The mineralogy and distribution of major and trace metals within the Lucky Strike hydrothermal deposits are not uniform, suggesting vent field scale variations in vent fluid composition (Bogdanov et al., 2006;Chavagnac et al., 2018). Both spatial and temporal variations in fluid chlorinity and CO 2 concentrations have been documented, indicating subseafloor fluid phase separation and effects of magma replenishment, respectively (Chavagnac et al., 2018;Langmuir et al., 1997;Pester et al., 2012;Von Damm et al., 1998). Results from previous sulfur isotope analyses from the main field at Lucky Strike indicate typical δ 34 S values for mid-ocean ridge hosted hydrothermal systems of between ∼0 and 10‰ (Rouxel et al., 2004). In this study, we present new mineralogical, bulk geochemical, and in situ sulfur isotope data from hydrothermal samples (n = 23) collected during the 2011-2015 MoMARsat maintenance cruises of the EMSO-Azores observatory (European Multidisciplinary Seafloor and water-column Observatory) from several of the hydrothermal edifices at Lucky Strike (Blandin et al., 2013;Cannat et al., 2011;Cannat & Sarradin, 2012;Sarradin & Cannat, 2014. We document the spatial variations in composition (mineralogical, geochemical, and in S isotopes) of the hydrothermal deposits to (a) evaluate the mineralogical controls on minor and trace element partitioning within the deposits; and (b) link these variations to the geological controls on hydrothermal venting conditions at the vent field scale.
In this study we also further investigate the influence of the Azores hotspot on the composition of hydrothermal deposits along the MAR. We develop geochemical criteria based on major and trace element compositions of hydrothermal deposits from MAR-hosted vent fields to fingerprint specific host rock compositions, which can vary from E-MORB to ultramafic rock.
Insights into the local and regional geological controls on the compositions of seafloor hydrothermal deposits is becoming increasingly important due to the growing interest in the economic potential of these deposits. Large seafloor hydrothermal deposits with high concentrations of base and precious metals (e.g., Cu, Zn, Au, and Ag) are potential targets for future mining (Hannington et al., 2011;Petersen et al., 2016;Rona, 2003). However, environmentally deleterious metals such as As, Cd, Se, Sb, which occur as trace metals within the sulfide minerals can also accumulate (Fallon et al., 2019;Hannington et al., 1999;Layton-Matthews et al., 2008;Martin et al., 2019). Results from this study will contribute to the understanding of the controls of substrate composition on the metal endowment, and therefore provide a framework for predictive assessment of the economic potential and associated environmental risks of exploiting seafloor massive sulfide deposits (SMS) based on geological setting on mid-ocean ridges.

Geological Setting
The Lucky Strike segment is located along the slow-spreading Mid-Atlantic Ridge, approximately 360 km southwest of the Azores archipelago, near the triple junction between the North American, African, and Eurasian plates (Figure 1; Cannat et al., 1999;Escartín et al., 2001). This segment is currently spreading at a full rate of ∼20-25 mm/yr (Argus et al., 2011;Cannat et al., 1999), and is bound to the north by a non-transform offset with the Menez Gwen segment and to the south by a non-transform offset with the North Famous segment (Figure 1). The Lucky Strike segment formed as a product of rifting of an oceanic plateau associated with the Azores hotspot that began at ∼10 Ma (Cannat et al., 1999;Escartín et al., 2001;Gente et al., 2003). The melt anomaly associated with the Azores hotspot is propagating southward, along the ridge, forming a V-shaped ridge that indicates shallowing of the ridges and its flanks toward the Azores hotspot, which is interpreted based on gravity and bathymetric data as a temporal and spatial variations in melt supply to the ridge axis that result from ridge-hotspot interactions (Cannat et al., 1999;Escartín et al., 2001).
The Lucky Strike segment is ∼70 km long, with a prominent axial volcano, named the Lucky Strike Seamount, that rises to 1500 mbsl in the central part of the segment from bathymetric lows of ∼3500 mbsl toward the northern end of the segment and ∼3200 mbsl toward the southern end of the segment, and is bounded by an overall ∼20 km wide fault-bounded rift valley (Figure 2). Bathymetry and sidescan sonar imagery record cyclic volcanic (magmatic) crustal construction followed by tectonic (rifting) phases . Recent volcanic episodes produced two volcanic centers that are aligned along the ridge axis and separated by a flat depression that is interpreted to be a recent but now fossil lava lake (Figure 3a; Humphris et al., 2002;Ondréas et al., 2009). Both volcanic structures are now cut by a series of ridge-parallel normal faults, with the older, northern, volcanic SÁNCHEZ-MORA ET AL.

10.1029/2021GC010073
3 of 24 edifice experiencing a higher degree of rifting than the younger, southern, volcanic edifice ( Figure 3a; Escartín et al., 2014;Humphris et al., 2002). The Lucky Strike rift valley floor is paved by volcanic rocks that range in composition from transitional mid-ocean ridge basalt (T-MORB) to E-MORB (Gale et al., 2011). Basalts with a more E-MORB affinity are found toward the center of the segment, whereas the T-MORB compositions are distributed along the rest of the segment. The stratigraphy of the lava flows indicate that E-MORB lavas are crosscut by T-MORB lavas (Gale et al., 2011;Langmuir et al., 1997). Lava morphologies vary from sheeted flows to pillow lavas, with sheeted flows dominant in the neovolcanic zone that hosts the hydrothermal fields Gini et al., 2021).
Focused hydrothermal activity is restricted to the central region of the Lucky Strike Seamount (Figure 3). Seismic reflection studies have identified an axial magma chamber that lies ∼3.5 km below the seafloor, and normal faults that propagate below the axial volcano (Combier et al., 2015;Singh et al., 2006). Microseismicity data suggest that hydrothermal circulation occurs primarily along-axis, and extends into the crust to within a few hundred meters of the axial magma chamber reflector in the narrow faulted area at the volcano summit (Crawford et al., 2013). The location of the vent field is also associated with a crustal magnetic low caused by hydrothermal alteration of the underlying rocks (Miranda et al., 2005).
At the vent field scale venting and hydrothermal deposits are associated with normal faults and occur for the most part near a fossil lava lake at the volcano summit ( Figure 3; Barreyre et al., 2012;Escartín et al., 2015;Fouquet et al., 1994;Ondréas et al., 2009). Focused venting also occurs 1.4 km to the east of the central ridge axis, at a site called Capelinhos . The spatial association of hydrothermal venting with faulting suggest that faulting is the primary control on fluid discharge (Barreyre et al., 2012;Escartín et al., 2015). Diffuse venting has been documented at Ewan, located 1.5 km south of the main field ( Figure 3; Escartín et al., 2015). The maximum temperature measured is 340°C at the main Lucky Strike field (South Crystal) and 324°C at Capelinhos (Barreyre et al., 2014;Chavagnac et al., 2018). The hydrothermal deposits at Lucky Strike were identified in previous studies via dredging, followed by autonomous underwater vehicle (AUV) and remotely operated vehicle (ROV) bathymetric mapping coupled with photomosaic imaging (Barreyre et al., 2012;Escartín et al., 2015;Langmuir et al., 1992Langmuir et al., , 1997Ondréas et al., 2009). Observations from time series imaging of the seafloor, acquired in 1996, 2006, 2008, and 2009, suggest that the heat flux associated with hydrothermal venting is decreasing, likely as a result of cooling of the axial magma chamber that is driving the fluid circulation in this system (Barreyre et al., 2012). Radioisotope ( 226 Ra) dating of hydrothermal barite indicates that hydrothermal venting at this site has been ongoing for at least 6,600 years and has accumulated at an average rate of ∼194 t/yr (Sánchez-Mora et al., submitted). Rock samples for this study were collected using the Victor 6000 ROV during the MoMARsat cruises from 2011 to 2015 on the R/V Pourquoi pas? and R/V Thalassa in 2012. Sampling mainly focused on hydrothermal deposits on several sites within the main vent field as well as on Capelinhos (Table 1). Samples for the main field at Lucky Strike (n = 19) and Capelinhos (n = 4) range from sulfide blocks at the bases of hydrothermal edifices to active and inactive sulfide-rich chimneys and samples that precipitated on temperature probes (Table 1).

Bulk Geochemistry
For bulk multi-element geochemistry, 23 samples were crushed at Memorial University using a ring pulverizer with a tungsten carbide grinding container and then analyzed at Activation Laboratories LTD (Actlabs). Samples were analyzed using: (a) Instrumental neutron activation analysis (INAA) for Au, Ag, As, Ba, Co, Fe, Na, Sb, Se, and Zn; and (b) Na 2 O 2 fusion and inductively coupled plasma optical emission spectroscopy (ICP-OES) and mass spectrometry (ICP-MS) for Ca, Mo, Ni, Sr, Al, Cd, Cu, Ga, Ge, In, Mg, Mn, Pb, S, Si, Sn, Tl, and V (Hoffman, 1992). Accuracy and precision for INAA values are both better than ±5%, based on repeat (n = 7) analysis of the GXR-1 standard. For the Na 2 O 2 fusion ICP method, accuracy is better than ±7%, based on repeat (n = 3) measurement of the OREAS 922 standard, with the exceptions of Ni and Sr, for which accuracy is 17%. Precision for the ICP method is better than 10%, except for Ni and Pb, which have precisions of 20% and 12%, respectively.  (Cannat et al., 1999;Escartín et al., 2001).

In Situ Sulfur Isotope Measurements
In situ sulfur isotope measurements were performed using a Cameca IMS 4f Secondary Ion Mass Spectrometer at the MAF-IIC Microanalysis Facility at Memorial University. Sulfide-bearing samples (n = 16) were embedded  (Barreyre et al., 2014;Escartín et al., 2015). The dashed red lines show the volcanic rifted structures v2 (higher degree of rifting) to the north and v1 in the south. White dashed outline is a fossil lava lake. (b) Venting of 222°C fluid (Tmax); (Von Damm et al., 1998); at a sulfide chimney at the Sintra vent complex within the main Lucky Strike hydrothermal field. The length of the robotic arm is ∼75 cm. (c) Chimneys and high-temperature venting at the Capelinhos site (Chavagnac et al., 2018). The temperature probe is ∼1 m long. Bathymetry sources: Ondréas et al. (2009) and Escartín et al. (2015 in epoxy in a 25.4 mm diameter aluminum ring and prepared as polished mounts, and sputter coated with 300 Å of Au to mitigate charging under primary ion bombardment. Isotopic analyses (n = 52) were performed by bombarding the sample with a 0.8-1.0 nA primary ion microbeam accelerated through a 10 keV potential and focused into a 15-20 µm diameter spot. Each spot was pre-sputtered for 120 s using a 10 μm raster to exclude exotic material in the polished surface from the analysis. Negatively charged sputtered secondary ions were accelerated into the mass spectrometer through a 4.5 keV potential. Signals for 32 S -, 34 Sand a background position at 31.67 Da were obtained by cyclical magnetic peak switching. Standard counting time and peak sequence used were 0.5 s at the background position, 2.0 s on 32 S -, and 6.0 s on 34 S -. A typical analysis consisted of accumulating 80 peak cycles. All peak signals were collected with an ETP 133H multiple-dynode electron multiplier (em) and processed through ECL-based pulse-counting electronics with an overall dead time of 11 ns. The production and detection of sputtered secondary ions produces an instrumental mass fractionation (IMF) bias between the actual 34 S/ 32 S of the sample and that measured by the mass spectrometer. The magnitude of the IMF varies substantially between sulfide minerals. For this reason, the 34 S/ 32 S measured in samples of pyrite were corrected for IMF by comparison to replicate measurements of in-house reference materials UL9B (pyrite; δ 34 S: +16.3‰) and a Norilsk chalcopyrite (δ 34 S: +8.4‰). Measured 34 S/ 32 S ratios are transformed to the Vienna Cañon Diablo Troilite (V-CDT) scale using 34 S/ 32 SVCDT = 0.0441626 (Ding et al., 2001). Data are presented in standard delta notation: where R = 34 S/ 32 S. Analyses yield internal precisions on individual δ 34 S determinations of better than ±0.3‰ (1σ) and the overall reproducibility, based on replicate standards analyses, is typically better than ±0.45‰ (1σ).

Mineralogy
Rock samples from Lucky Strike were collected primarily from the hydrothermally active Sintra, Tour Eiffel, Chimiste, Isabel, White Castle and Bairro Alto edifices ( Figure 3a; Ondréas et al., 2009). Within the samples, mineral distributions change from the exterior to the interior of the chimneys, as typically described at black smoker deposits (Fouquet et al., 2010;Hannington et al., 1995;Haymon, 1983), with a lower temperature mineral assemblage of barite, anhydrite, marcasite, and pyrite with millimeter-scale goethite rims dominating the exterior of the chimney, and a higher-temperature mineral assemblage of sphalerite and chalcopyrite dominating the interior and lining fluid conduits (Figures 4a-4e). Early marcasite forms plumose, colloform, or ringlike textures. Later stage marcasite is massive and euhedral. Subhedral to euhedral pyrite overgrows marcasite and co-precipitates with sphalerite that exhibits a range of textures, from plumose to subhedral and massive (Figures 4d, 4f and 4g). Late stage chalcopyrite has massive and euhedral textures, and is often replacing or overgrowing sphalerite (Figures 4d, 4h and 4i). Supergene alteration consists of exterior goethite rinds and atacamite (Figure 4f), and secondary covellite and bornite replacing chalcopyrite (Figure 4g). No major mineralogical differences are identified between Sintra, Tour Eiffel, and White Castle, except for an absence of pyrite in samples collected from Sintra, and absence of covellite in samples from White Castle.
Samples from Capelinhos show similar mineralogical assemblages and textural characteristics to the main Lucky Strike field, with lower temperature assemblages in the outer chimney walls that transition to higher temperature assemblages toward the interior. However, the primary difference is the occurrence of a second generation of sphalerite (Sph2), chalcopyrite (Cpy2), and barite (Ba2), that overprint previous generations (Figures 4g, 4h and 4i) suggesting temporal fluctuations in temperature of mineralizing fluids (Eldridge et al., 1983). Supergene alteration includes replacement of chalcopyrite by covellite and minor bornite, goethite and atacamite.  Principal component analysis (PCA) was performed to identify statistical trends in the chimney sample data set. The analysis was performed using the 28 elements of which at least 50% of the samples contained concentrations greater than their respective detection limits. When reported concentrations were below detection limits, a value representing 65% of the detection limit was used (Palarea-Albaladejo et al., 2014). A sensitivity analysis using different thresholds for percentages of samples below detection limit and different values to represent the analyses below detection limit resulted in negligible changes to the results of the PCA analysis. A centered-log-ratio transformation was applied to the raw data to avoid closure problems and spurious correlations (Pawlowsky-Glahn & Egozcue, 2006). The centered-log-ratio transformation (clr(x)) was calculated using the following equation:

Bulk Geochemistry of Hydrothermal Deposits
where D = the number of samples.
The PCA ( Figure 5) reveals a cluster of elements associated with high-temperature sulfide minerals (e.g., Co, Cu, Se and Sn) with a negative PC1 loading, and a second cluster of elements associated with lower-temperature sulfide minerals (Ga, Sb, Zn, Cd, Mn, Si) with a positive PC1 loading. Principal component 2 has positive loadings for sulfide mineral associated elements (transition and post-transition metals), and negative loadings for sulphate and oxide minerals associated elements (alkaline earth metals such as Ba, Sr, Ca, and Mg) but also Ni. Principal component 1 accounts for 37% of the variability in the data set and PC2 accounts for 17% of the variability. Table 2 summarizes results from in situ sulfur isotope measurements for marcasite and chalcopyrite (Figures 6  and 7). The δ 34 S values for marcasite and pyrite (n = 30) average 1.0‰ (1σ = 1.8) and vary from −2.5 to 5.2‰.

of 24
Values for chalcopyrite (n = 22) average 4.4‰ (1σ = 2.1) and range from 1.0 to 8.7‰. Notable low (negative) δ 34 S values of −2.5 and −2.4‰ where obtained on marcasite from the Y3 and Isabel sites, respectively, which are part of the main field. Average marcasite and chalcopyrite sulfur isotope compositions from the main field (0.3‰, n = 24, and 3.3‰, n = 15, respectively) are lower than those for Capelinhos (3.7‰, n = 6, and 6.8‰, n = 7, respectively; Figure 7). Marcasite δ 34 S values do not vary significantly as a function of texture at the main field ( Figure 6). At both sites, chalcopyrite is, on average, isotopically heavier by ∼3‰ than marcasite (Figure 7). There is no apparent correlation between δ 34 S values and age for the seventeen samples for which both S isotope and age data are available (Sánchez-Mora et al., submitted).

Discussion
The bulk, minor and trace element geochemical data presented in this study expands on the data originally presented by Bogdanov et al. (2006) by including a more extensive suite of elements such as Au, Ag, As, Ba, Mo, Sb, Se, Sb, Si, and Sn (see Table 1). Additionally, this study includes data from samples from other areas of the vent field not included in the study by Bogdanov et al. (2006), including the Capelinhos site, for which no data has been published to date. Results from the principal component analysis reflect the strong temperature control and influence of seawater on the mineral associations and trace element distributions within the deposits ( Figure 5). An analysis of the distribution of the different sites within the Lucky Strike vent field that occur at different water depth within the field (Table 1) over the PC1 and PC2 space contrast with the interpretation of Bogdanov et al. (2006), who suggested that, due to phase separation occurring at the seafloor or in the shallow subseafloor, the highest temperature fluids occur at the deepest vents within the vent field. This phase separation was hypothesized to be a main control on the variability of composition of the hydrothermal deposits (Bogdanov et al., 2006). However, our data indicates no systematic variation in bulk geochemistry with water depth or deposit age (∼6,600 years to present; Sánchez-Mora et al., submitted). Therefore, differences in mineral assemblages and geochemistry likely only reflects the stage (low-or high-temperature) at which the sample was formed and not necessarily linked to phase separation.  14 of 24

Controls on Minor and Trace Element Distributions at Lucky Strike
Minor and trace elements occur within the main mineral phases (chalcopyrite, marcasite, pyrite and sphalerite) in hydrothermal sulfide deposits and can be incorporated into their crystal structure or can occur as micro-or nanoscale inclusions (Cook et al., 2016;Fontboté et al., 2017;Monecke et al., 2016). Minor and trace element distribution is generally evaluated via microanalytical techniques such as an electron microprobe and/or laser ablation ICP-MS (e.g., Berkenbosch et al., 2019;Grant et al., 2018;Keith et al., 2016;Melekestseva et al., 2017). Our approach uses PCA on bulk geochemistry of hydrothermal samples from the Lucky Strike vent field to determine the minor and trace element variability, which we compared to previous geochemical work conducted at Lucky Strike by Bogdanov et al. (2006) and compared to other similar studies on mid-ocean ridges.
The mineralogy of the Lucky Strike hydrothermal deposits are typical for basalt hosted mid-ocean ridge deposits, which is generally dominated by marcasite, pyrite, sphalerite, and chalcopyrite with the notable exception of the high abundance of barite, which has been linked to high Ba concentrations in the underlying E-MORB substrate (Langmuir et al., 1992). High-temperature interiors of chimneys are dominated by chalcopyrite and euhedral pyrite and marcasite. Concentrations of In, Se, Co and Sn correlate with the major mineral-forming elements (Cu, Fe, and S) that comprise this high-temperature mineral assemblage ( Figure 5). These associations have been documented in other seafloor hydrothermal sites (e.g., Auclair et al., 1987;Hannington et al., 1991). Cobalt has been previously documented to be concentrated in marcasite and pyrite at Lucky Strike (Bogdanov et al., 2006). The correlation between Co with Cu and Fe that are associated with the precipitation of minerals at high-temperatures suggests that Co is hosted in euhedral pyrite and marcasite that precipitated associated with these high-temperature mineral assemblages ( Figure 5). At the basalt-hosted TAG hydrothermal field further south on the MAR, In is hosted primarily as a trace metal within chalcopyrite and marcasite, Sn is largely hosted within chalcopyrite, and Se is largely hosted in chalcopyrite and pyrite to a lesser extent (Grant et al., 2018). Similarly, Lein et al. (2010) and Bogdanov et al. (2008) report high concentrations of In within high-temperature mineral assemblages at Menez Gwen and Broken Spur, respectively, both also located at the MAR to the North and South of Lucky Strike, respectively. Selenium has also been documented to occur primarily within chalcopyrite at Lucky Strike (Rouxel et al., 2004) but also in pyrite in a range of other seafloor hydrothermal deposits (e.g., Monecke et al., 2016, and references therein).
The lower temperature exteriors of chimneys from Lucky Strike are dominated by sphalerite (ZnS), barite (BaSO 4 ), plumose marcasite and amorphous silica. Silver, Au, As, Ga, Cd, Sb, Tl, Pb, and Ge abundances correlate with the major mineral-forming elements (Zn, Ba, and Si) of this lower temperature mineral suite. The association of some of these trace elements (As, Sb, Cd, Mn, and Ag) with lower temperature minerals in the PCA ( Figure 5) such as marcasite and sphalerite has been documented at Lucky Strike (Bogdanov et al., 2006) as well as at other hydrothermal deposits hosted on the MAR, including Menez Gwen (Lein et al., 2010), TAG (Grant et al., 2018), Broken Spur (Bogdanov et al., 2008). The elements Ca, Ba, and Sr, which are associated with the sulphate minerals anhydrite and barite, have a distinct negative loading with respect to PC2. Barium and Sr have minor positive loadings and Ca has a minor negative loading with respect to PC1 ( Figure 5). The minor negative loading of Ca with respect to PC1 is consistent with anhydrite occurring primarily in association with the high-temperature sulfide mineral suite due to its absence at lower temperatures due to its retrograde solubility at temperatures below 150°C (Blount & Dickson, 1969). In contrast, Ba and Sr have a positive loading with respect to PC1, consistent with the typical association of barite with lower-temperature sulfide minerals . The proximity of Sr to Ba in Figure 5 is likely the result of barite being significantly more abundant than anhydrite at Lucky Strike, as indicated by the higher Ba concentrations (and higher barite contents) in the samples compared to Ca.
Magnesium, V, Ni, and to a lesser extent, Mo and Na have a negative loading with respect to PC1 (Figure 5). At TAG, redox-sensitive elements such as V, Ni, and Mo are associated with the precipitation of pyrite and marcasite that has interacted with cold, Na-and Mg-rich seawater toward the exteriors of vents (Grant et al., 2018). Alternatively, Ni, and V can also be associated with the formation and scavenging of Fe-oxyhydroxides (German & Seyfried, 2014). Regardless, the negative loading with respect to PC2 discriminates elements associated with sulphate and oxide minerals that precipitate upon direct interactions with seawater, which is consistent with the three samples with the lowest PC2 loadings that are barite and Fe-oxyhydroxides rich.

Differences in S Isotope Compositions Between the Main Lucky Strike Field and Capelinhos
Overall, the sulfur isotope compositions of the sulfide deposits at Lucky Strike range between −2.5 and 8.7‰ (Figure 7), which is within the range of average δ 34 S values for sulfide minerals at other sediment-free and basalt-hosted mid-ocean ridge hydrothermal deposits (Hannington et al., 2005;Zeng et al., 2017). The range in δ 34 S values for the main field reported in this study are similar to those reported by Rouxel et al. (2004). The mineralogy of the Capelinhos site is similar to the mineralogy of the deposits at the main field, which is consistent with evidence from vent fluid trace element concentrations that correlate with chlorinity and indicate a common hydrothermal fluid source for these sites (Chavagnac et al., 2018). However, the S isotope compositions of sulfide minerals from Capelinhos are ∼3.5‰ higher than the average values at the main field (Figure 7). The range at Capelinhos is comparable to other sediment-free fast to slow spreading mid-ocean ridges sites such as TAG, East Pacific Rise South, and Axial Seamount (Hannington et al., 2005;Zeng et al., 2017).
At mid-ocean ridge-hosted hydrothermal sites, δ 34 S values of between ∼0 and 10‰ are usually interpreted to be a result of two component mixing between igneous-derived sulfur (δ 34 S ≈ 0‰) leached from mid-ocean ridge crust and seawater sulphate (δ 34 S = 21‰); (Hannington et al., 2005;Shanks, 2001;Zeng et al., 2017). It is possible that the more positive δ 34 S values for sulfide minerals at Capelinhos are due to a higher relative contribution of reduced sulfur derived from seawater sulphate or a longer fluid flow pathway feeding the Capelinhos vents, or a combination of both (Chavagnac et al., 2018;Escartín et al., 2015). Also, in sediment free environments, the addition of isotopically-light H 2 S associated with disproportionation of magmatic SO 2 can also drive the overall system toward lower δ 34 S values (Gamo et al., 1997;McDermott et al., 2015;Shanks, 2001). Fluctuations in vent fluid CO 2 content (Pester et al., 2012) and recent magmatic intrusions detected from seismicity at Lucky Strike (Dziak et al., 2004) indicate that magmatic volatiles may contribute to the hydrothermal system. A higher proportion of such volatiles could result in the relatively lower δ 34 S values in the hydrothermal precipitates in the main field. However, CO 2 fluid data is not available for Capelinhos to assess differences in volatile input between this site and the main field. In addition, there is no mineralogical evidence of direct magmatic volatile input, such as the presence of high-sulfidation alteration assemblages, sulfosalts, and elevated concentrations of volatile associated trace elements such as Bi, Se, and Te (de Ronde et al., 2011;Martin et al., 2019). Rouxel et al. (2004) speculate that, at Lucky Strike, there is also a component of a fractionated sulfur source associated with microbial activity below the fossil lava lake. These interpretations highlight the ambiguity inherent in the interpretation of δ 34 S values on their own and the requirement of additional information to assess processes driving different sulfur isotope compositions at Lucky Strike.
Vent fluids at the main field are also depleted in Fe, enriched in Cl, and contain higher 87 Sr/ 86 Sr, relative to Capelinhos fluids (Chavagnac et al., 2018). Chavagnac et al. (2018) interpret the higher 87 Sr/ 86 Sr at the main field to indicate higher rock/water ratios, even though the field lies directly above the heat source and the overall flow pathway is likely shorter. Longer residence times for fluids venting at the main field may result from locally reduced permeability. The fossil lava lake may act as a hydrological cap for ascending fluids, promoting lower 10.1029/2021GC010073 16 of 24 overall permeabilities in the subsurface (Arnulf et al., 2014) leading to increased interaction with altered basalts, thus driving the δ 34 S values of the fluids and deposits closer to zero (Figure 8). Chavagnac et al. (2018) report a ∼65% depletion in the dissolved Fe concentration of the fluids from the main field, relative to Capelinhos. Concentrations of other dissolved species (e.g., Mg, Mn, Na, and SO 4 ) and vent fluid pH indicate that this depletion is not a result of seawater dilution. Therefore it is suggested that significant amounts of precipitation of Fe-bearing minerals occur in the subseafloor below the main field (Chavagnac et al., 2018). The precipitation of pyrite/marcasite would result in a temperature-dependent S isotope fractionation between the precipitated pyrite/marcasite and reduced S in the fluid. The fraction of reduced S removed from the fluid due to subseafloor pyrite/marcasite precipitation can be modeled over a range of temperatures as an open system Rayleigh distillation: where R is the isotopic ratio of the fraction f of reduced sulfur remaining in the fluid after pyrite/marcasite precipitation, R 0 is the initial sulfur isotope ratio of the fluid, and α is the fractionation factor (α = R Py /R H2S ). Compared to the vent fluids at the main field, the composition of the endmember venting fluids at Capelinhos have been interpreted to be more representative of the fluids within the deeper part of the hydrothermal reaction zone, based on the linear correlation of trace elements and chlorinity in the hydrothermal fluids at Lucky Strike (Chavagnac et al., 2018;Pester et al., 2012). The δ 34 S values from sulfide minerals at Capelinhos can therefore be used to represent R 0 (i.e., the isotopic composition of the ascending fluids below the main field) and the δ 34 S values from Figure 8. Summary sketch of the location and faulting at the Lucky Strike hydrothermal field. The stockwork is interpreted based on the sulfur isotope data from this study. Axial magma chamber (∼3.5 km depth) outline from Combier et al. (2015) and layer 2A from Seher et al. (2010), both derived from seismic reflection data.
the main field represent R. The temperature dependent fractionation factors (α), determine the different mass fractions (ƒ) at which the ∼3.5‰ difference between Capelinhos and main Lucky Strike can be obtained. If fractionation factors are used between H 2 S and pyrite (Ohmoto & Rye, 1979) from 150 to 350°C, fractions of 0.21 to 0.034 are obtained, respectively (Hannington, 2014).
Results from sulfur isotope modeling suggest that at least ∼80% of the original H 2 S in the hydrothermal fluid at the main field precipitated below the seafloor, dominantly as pyrite/marcasite (Figure 8). A similar proportion (75%) of major and trace element deposition below the seafloor has been estimated at the seawater-dominated Reykjanes geothermal system in Iceland, which is considered an analog to seafloor systems (Grant et al., 2020). The average value of H 2 S in hydrothermal fluids at the main field vents is 2.7 mmol/kg (Charlou et al., 2000). If this value represents less than 20% of the original H 2 S concentration, the original fluid H 2 S concentration would have been greater than ∼14 mmol/kg. The amount of reduced S trapped in the subsurface as pyrite/marcasite is reasonably consistent with the amount of Fe estimated to have been trapped in the subsurface based on differences in vent fluid chemistry between the main field and Capelinhos (Chavagnac et al., 2018).
Chalcopyrite is another Fe-bearing mineral that could have precipitated and trapped both Fe and S in the subsurface and commonly forms in the upflow zones of seafloor hydrothermal systems (Franklin et al., 2005;Galley et al., 2007). However, the fractionation factors between chalcopyrite and H 2 S at subseafloor hydrothermal conditions are extremely low (e.g., <0.3 (1000lnα) at 150°C, Ohmoto & Rye, 1979) and very large amounts of reduced sulfur would have to precipitate in the subsurface to generate the predicted fractionations, requiring initial fluid H 2 S concentrations of >300,000 mmol. For comparison, maximum reported H 2 S concentrations for vent fluids are on the order of 110 mmol (Von Damm et al., 1995). Therefore, although chalcopyrite may be precipitating in the subseafloor, the isotopic differences between Capelinhos and the main field suggest that pyrite is the primary sulfide phase being precipitated. This interpretation is consistent with evidence from subsea drilling at TAG and from VMS deposits that show that upflow zones can be enriched in chalcopyrite but are dominated by pyrite (Galley et al., 2007;Knott et al., 1998).
It should be noted that the calculations and results presented above are predicated assuming isotopic equilibrium between the vent fluid and minerals. In seafloor hydrothermal systems, isotopic equilibrium can be evaluated comparing known temperatures of formation to temperatures calculated between mineral pairs-based equilibrium sulfur isotope fractionation. For example, there is a uniform difference in isotopic composition of ∼3‰ between neighboring chalcopyrite and marcasite crystals at both the main field and Capelinhos ( Figure 6). This difference is independent of sample location, and depositional age as there is no significant variation in δ 34 S values and mineral textures (e.g., plumose vs. euhedral marcasite) (Sánchez-Mora et al., submitted). The isotopic enrichment in 34 S in chalcopyrite, relative to marcasite, is consistent with equilibrium fractionation processes. Under equilibrium conditions, a 3‰ fractionation between chalcopyrite and marcasite corresponds to a crystallization temperature of 114°C (using fractionation factors by Kajiwara & Krouse, 1971). This temperature is likely too low for typical crystallization temperatures for chalcopyrite (>250°C) although marcasite can form at temperatures of up to 240°C under hydrothermal conditions (Hannington et al., 1995;Murowchick & Barnes, 1986), suggesting that these mineral phases were not in isotopic equilibrium at the time of formation. This is consistent with textural evidence that indicates that these mineral phases did not co-precipitate (Figure 6), and the difference in isotopic composition between the two mineral phases is likely a result of either equilibrium or non-equilibrium isotopic partitioning between each mineral phase and the hydrothermal fluid at different temperatures of mineral formation and/or other possible kinetic isotopic effects, such as effects of microbial processes during marcasite formation (Juniper et al., 1988). Overall, an important outcome from this analysis is that subsurface sulfide mineral precipitation can change the S isotopic composition of the vent fluid and surficial deposits (Figure 8), yielding surficial isotopic compositions that are indistinguishable from isotopic compositions that result from simple two-component mixing between mantle sulfur and reduced seawater sulphate (Ohmoto & Lasaga, 1982;Ono et al., 2007).

Geological Controls on the Composition of Lucky Strike and Other Vent Fields Along the Mid-Atlantic Ridge
The most distinctive compositional characteristic of both Lucky Strike and Menez Gwen (a vent field located ∼90 km northeast from Lucky Strike, closer to the Azores hotspot) is the abundance of barite (Bogdanov et al., 2005;Lein et al., 2010). Compared to other basalt or ultramafic-hosted vent fields along the MAR that are 18 of 24 not hotspot influenced, such as TAG, Snake Pit, and Rainbow, Lucky Strike contains significantly more barite, and higher concentrations of Ba, Sr and Mo, but lower Au and Sn ( Figure 9); (Bogdanov et al., 2002;Fouquet et al., 1993;Grant et al., 2018;Honnorez et al., 1990;Krasnov et al., 1995;Marques et al., 2007). The basaltic substrate at Lucky Strike and Menez Gwen is dominated by E-MORB (Gale et al., 2011;Langmuir et al., 1997), which is enriched in incompatible elements (e.g., Ba and Sr) compared to the normal mid-ocean ridge basalts (N-MORB) that host TAG and Snake Pit (Hannington et al., 2005). The E-MORB substrate composition and associated enrichment in incompatible elements (Ba and Sr) is linked to the melting of the metasomatized mantle associated with the Azores hotspot (Langmuir et al., 1997).
In contrast to E-MORB hosted Lucky Strike and Menez Gwen, the active TAG mound, which contains relatively low concentrations of Ba and Sr, is hosted within N-MORB in an active detachment fault zone (Humphris et al., 2015). Rainbow is hosted in ultramafic rocks but is influenced by basaltic rocks in the vicinity of the hydrothermal site. Rainbow has a relatively high Ni and Co content, which is typical for ultramafic-hosted sites (Marques et al., 2007;Mozgova et al., 2008). At ultramafic sites, serpentinization causes the release of Co and Ni during the alteration of primary silicate minerals such as olivine (Marques et al., 2007). Barium enrichment in E-MORB hosted hydrothermal sites and Co enrichment in ultramafic hosted hydrothermal sites have been widely documented (e.g., Hannington et al., 2005). The Ba/Co ratio within hydrothermal deposits can therefore be used to discriminate E-MORB hosted deposits from deposits hosted on other mafic and ultramafic substrates at sites where the composition of the host rock is not known (Figure 10a). To test the utility of this ratio, a compilation of four groups of substrate types were compiled (Table S2) for which bulk geochemistry of hydrothermal deposits was available: (a) E-MORB (Lucky Strike, Menez Gwen, and Endeavor); (b) N-MORB (TAG, Snake Pit, and Broken Spur); (c) ultramafic (Rainbow and Logatchev); and (d) mixed mafic and ultramafic (Beebe, Kairei, Yuhuang-1, and Daxi; Figures 10a and 10b). The median Ba/Co values are 180, 3, 3, and 0.006 for E-MORB, N-MORB, ultramafic, and mixed mafic/ultramafic groups, respectively ( Figure 10b). However, concentrations of Ba and Co can also be significantly affected by temperature during mineral precipitation, with Co associated with high-temperature mineral assemblages and Ba associated with lower-temperature mineral assemblages ( Figure 5). Therefore, to test whether changes in Ba/Co values primarily reflect substrate composition or temperature, the ratios are plotted against Cu/Zn, another geochemical proxy for temperature, for the different substrate groups ( Figure 10). Elevated Cu values are associated with precipitation of high-temperature chalcopyrite (250-350°C) and high Zn values are associated with precipitation of sphalerite at lower temperatures (<250°C; Figure  5; Hannington, 2014). For the E-MORB group, the Ba/Co values show a general negative trend, relative to Cu/ Zn, which reflects the temperature dependence of both ratios within a group. However, when comparing data Figure 9. Average bulk geochemistry of hydrothermal samples of selected Mid-Atlantic Ridge hosted hydrothermal sites, normalized to average mid-ocean ridge basalt composition (Arevalo & McDonough, 2010). The diamond symbol indicates sites hosted in detachment fault systems. Lucky Strike data is from this study and Bogdanov et al. (2006), Menez Gwen (Bogdanov et al., 2005), TAG (Hannington, 1993;Hannington et al., 1991;Krasnov et al., 1995;Lisitsyn et al., 1990;Rona et al., 1993;Thompson et al., 1988), Snake Pit (Fouquet et al., 1993;Hannington et al., 1991;Honnorez et al., 1990;Kase et al., 1990;Krasnov et al., 1995), and Rainbow (Bogdanov et al., 2002;Marques et al., 2007). Menez Gwen (n = 12), Lucky Strike (n = 50), Rainbow (n = 6), TAG (n = 190), and Snake Pit (n = 105). Full data set provided in Table S2. Figure 10. (a) Ratios from selected seafloor massive sulfide deposits (SMS) deposits grouped according to host rock in mid-ocean ridges. Ba/Co versus Cu/Zn. Ba/ Co ratio used as a proxy for proximity to a hotspot, Cu/Zn used as a proxy of temperature of formation. (b) Box and whisker plots of quartiles for Ba/Co and Cu/Zn (boxes with black line for median, black dot for mean, and number of samples). Enriched mid-ocean basalts (E-MORB) includes data from Lucky Strike (this study and Bogdanov et al. (2006)), Menez Gwen (Bogdanov et al., 2005), and Endeavor (Morgan & Selk, 1984;Samson, 1986;Tivey & Delaney, 1986;Tivey et al., 1999;Toffolo et al., 2020). N-MORB includes TAG (Hannington, 1993;Hannington et al., 1991;Krasnov et al., 1995;Lisitsyn et al., 1990;Rona et al., 1993;Thompson et al., 1988), Snake Pit (Fouquet et al., 1993;Hannington et al., 1991;Honnorez et al., 1990;Kase et al., 1990;Krasnov et al., 1995;Toffolo et al., 2020), and Broken Spur (Bogdanov et al., 1995(Bogdanov et al., , 2008Lisitsyn et al., 1999;Peresypkin et al., 1999;Toffolo et al., 2020). Ultramafic-hosted includes Rainbow (Marques et al., 2007) and Logatchev (Bogdanov et al., 1997;Krasnov et al., 1995;Lisitsyn et al., 1999;Mozgova et al., 1999;Murphy & Meyer, 1998;Peresypkin et al., 1999). Mixed mafic and ultramafic includes Beebe (Webber et al., 2015), Kairei (Wang et al., 2014), Yuhuang-1 (Liao et al., 2018), and Daxi (Wang et al., 2021). Data in Table S2. 20 of 24 geochemical composition of the off-axis Capelinhos site is similar to the deposits from the main field. However, the sulfur isotopic composition of Capelinhos is heavier by ∼3.5‰ compared to the rest of Lucky Strike. These differences reflect variations in length and/or permeability in the pathways of the uprising hydrothermal fluid and distinct fluid/rock interactions at each of these sites, including evidence from S isotope modeling of extensive subseafloor marcasite/pyrite mineralization below the main site. Here, S isotopic data corroborate previously reported variations in fluid chemistry between different sites that indicate that >80% of the available H 2 S in the ascending hydrothermal fluid precipitates in a stockwork zone below the seafloor. These results highlight that the interpretation of S isotope compositions for hydrothermal deposits along mid-ocean ridges cannot necessarily be interpreted simply within the context of a two-component mixing model between mantle-and seawater-derived sulfur sources, and that subseafloor mineralization and isotopic exchange can also affect the isotopic compositions of surficial sulfide deposits.
The Ba/Co ratio of hydrothermal deposits provides a temperature-independent geochemical tool to discriminate the composition of the substrate beneath hydrothermal vent fields at mid-ocean ridges, especially E-MORB hosted site that have a distinct elevated Ba/Co ratio. High Ba/Co ratios at Lucky Strike and Menez Gwen indicate an E-MORB substrate associated with the Azores hotspot. In contrast, low Ba/Co ratios, are associated with hydrothermal sites hosted in N-MORB, ultramafic, and sites that have a mixed mafic and ultramafic substrates.

Data Availability Statement
All the geochemical data (bulk geochemistry and in situ sulfur isotope) is included in this paper and in the Supporting Information and can be also accessed at https://doi.org/10.5281/zenodo.6035032 (Sánchez-Mora et al., 2022).