Talc-dominated seafloor deposits reveal a new class of hydrothermal system

The Von Damm Vent Field (VDVF) is located on the flanks of the Mid-Cayman Spreading Centre, 13 km west of the axial rift, within a gabbro and peridotite basement. Unlike any other active vent field, hydrothermal precipitates at the VDVF comprise 85–90% by volume of the magnesium silicate mineral, talc. Hydrothermal fluids vent from a 3-m high, 1-m diameter chimney and other orifices at up to 215 °C with low metal concentrations, intermediate pH (5.8) and high concentrations (667 mmol kg−1) of chloride relative to seawater. Here we show that the VDVF vent fluid is generated by interaction of seawater with a mafic and ultramafic basement which precipitates talc on mixing with seawater. The heat flux at the VDVF is measured at 487±101 MW, comparable to the most powerful magma-driven hydrothermal systems known, and may represent a significant mode of off-axis oceanic crustal cooling not previously recognized or accounted for in global models.

H ydrothermal activity at mid-ocean ridges is dominated by basalt-hosted, high-temperature, metal-rich vent systems driven by magmatic activity 1 . However, the discovery of ultramafic-hosted hydrothermal vent fields (for example, the high-temperature Rainbow and the low-temperature Lost City Vent Fields) demonstrates the diversity of hydrothermal activity associated with medium-slow spreading ridges 2,3 . Tectonic exposure of upper mantle and lower crustal rocks gives rise to a more heterogeneous basement than at intermediate-fast spreading ridges, and detachment faulting provides pathways to enhance hydrothermal circulation 4,5 .
A new class of hydrothermal system, the Von Damm Vent Field (VDVF), was discovered in the Caribbean during cruise JC044 of the RRS James Cook in April 2010, which hosts a community of hydrothermal vent fauna similar to those at the Mid-Atlantic Ridge 6 . The VDVF hydrothermal plume is rich in methane, has a significant Eh anomaly (indicative of reduced fluids) and is free of metallic particulates 6 . Here we describe how the unusual mineralogy of the VDVF results from mixing between moderate-temperature vent fluid and cold seawater. The processes leading to the formation of the VDVF may be widespread throughout medium to ultraslow mid-ocean spreading ridges medium-ultraslow spreading mid-ocean ridges and could play a significant role in the cooling and chemical exchange between oceanic crust and seawater.

Results
Geological setting. The VDVF is located on the western flanks of the ultraslow spreading Mid-Cayman Spreading Centre, the deepest spreading centre on Earth (Fig. 1a). It is situated at a depth of 2,280 mbsl, 13 km to the west of the rift axis 6,7 on the upper slopes of Mt Dent, an oceanic core complex (OCC) formed by detachment faulting (Fig. 1b). At this location, the basement age is estimated to be between 1 and 2 Ma (based on distance from spreading centre and spreading rate) and comprises meta-gabbro, dolerite dykes and serpentinized peridotites that are partially covered by calcareous pelagic sediment 7 . The hydrothermally active area of the VDVF comprises three overlapping conical-shaped talc mounds, up to 75 m high and 150 m in diameter, aligned north-northwest-south-southeast (NNW-SSE) (Fig. 1c). The summit of the largest mound (Main  Cone), which is the northern mound at the active site, hosts a 3-m tall, 1-m diameter chimney (The Spire, Fig. 1d). Hydrothermal fluid venting (215°C) from The Spire has low concentrations of particles and a pH of 5.8 (determined at standard temperature and pressure (STP)). A 1-m diameter orifice (Main Hole) located at the base of The Spire vents fluids of up to 91°C. Elsewhere across the VDVF, smaller orifices vent fluids of up to 138°C (for example, Hotter than Hole, Chimlets 1 and Chimlets 2- Fig. 1c). Talc rubble at the base of the Main Cone onlaps the surrounding calcareous sediment-a relationship reversed at the base of the southern mound, indicating an increase in the age of the mounds towards the south (Fig. 1c). A further series of hydrothermally inactive, conical-shaped talc mounds is located 700 m to the south and east of the three active mounds. These include a 90-m high, sediment-covered cone (Mystic Mountain), which has twice the volume of the Main Cone. Samples from these extinct mounds are similar in composition, mineralogy and texture to the active mounds. With an estimated sedimentation rate of between 2 and 5 cm ka À 1 (ref. 8), a thickness of 1 m or more of pelagic sediment covering most of Mystic Mountain indicates that its construction by hydrothermal activity ceased at least 20,000 years ago.
Petrology. Compared with hydrothermally active seafloor deposits elsewhere, the mounds and chimneys of the VDVF are highly unusual, constituting 85-90% talc by volume with up to 10% microcrystalline silica and 5% disseminated sulphides ( Supplementary Fig. 1). In hand-specimen, the hydrothermally active chimneys show millimetric layers of laminated and botryoidal talc (Fig. 2a). These layers are parallel and have an internal colloform structure (Fig. 2a). Under scanning electron microscope (SEM), broken surfaces reveal dendritic networks and botryoidal masses of talc, indicative of growth into open void spaces (Fig. 2b,c). Concentric bands of sulphide up to 50-mm thick are present as internal growth bands within the talc masses ( Fig. 2d) and microcrystalline silica infills pore spaces (Fig. 2e). Talus forming the flanks of the mounds has a similar mineralogy to the venting chimneys, except that the botryoidal texture and open pore spaces are largely replaced by massive fine-grained talc, and the associated sulphides are mostly oxidized. Contrary to initial reports 6 , talc is the dominant mineralogy at the VDVF; only two of 50 samples recovered contain trace amounts of anhydrite and gypsum. Together with its botryoidal form, zoning, and layering. The talc deposits indicate precipitation from a hydrothermal fluid. The petrology of the hydrothermally active chimneys indicates a paragenetic sequence comprising initial talc growth into open space with co-precipitation of minor sulphide, followed by infilling of the pore space by later-stage microcrystalline silica. Loss of sulphides and a reduction in base-metal content in the flank talus indicate dissolution following exposure to cold and oxygenated seawater.
Mineralogy. Bulk X-ray diffraction analyses (Supplementary Fig. 1) and microscopy of the VDVF chimney and mound material confirm the dominant presence of talc, with microcrystalline silica and sulphide constituting up to 15%. Chalcopyrite is the dominant sulphide (70%), indicative of episodically higher vent temperatures 9 , with the remainder being 20% pyrite, 5% sphalerite and 5% galena. The presence of sulphides at the VDVF, and a lack of sulphate in the end-member fluids, confirms a reducing environment for the precipitation of the chimney and mound talc deposits 10 .
Other studies of talc, recovered from seafloor environments elsewhere, have revealed significant amounts of intracrystalline layers of smectite clays reflecting the influence of sediment alteration and precipitation from Mg-rich pore waters 11 . X-ray diffraction analyses of unorientated air-dried mounts of talc separates from the VDVF chimneys and talus deposits show peaks at B9.6 Å (for the 001 plane) and B4.7 Å (Supplementary Fig. 2) confirming talc (Mg 3 Si 4 O 10 (OH) 2 ) as the dominant phase. Shifts in peaks of the 001 plane from 9.604-9.627 Å for the air-dried mounts to 9.339-9.525 Å for the glycolated mounts show the presence of up to 10% of a smectite clay that is interlayered within the crystalline talc structure. The presence of peaks at 1.529 and 1.532 Å (for the 060 plane) indicate a tri-octahedral structure for the smectite clay interlayers ( Supplementary Fig. 2). The dominance of talc (90%), with up to 10% clay interlayers, contrasts with previously reported, sediment-hosted seafloor talc deposits 11 .
Deposit geochemistry. The high proportions of talc and microcrystalline silica are reflected in the whole-rock geochemistry ( Table 1)  Rare earth element (REE) patterns for talc, separated from hydrothermally active chimneys and mound talus samples, have shallow U-shaped profiles dominated by a large and positive europium anomaly (Fig. 3). Enrichment in the light RREs (LREEs) is indicated by La (N) /Sm (N) (where N ¼ chondritenormalized values) ratios ranging between 0.9 and 10.4, with an average of 3.3 for chimneys and 3.1 for mound talus ( Table 2). The REE profiles also have slight enrichment in heavy RRE (HREE) with average Dy (N) /Yb (N) ratios of 0.8 for chimney material and 0.9 for the mound talus. The magnitude of the positive Eu anomaly, defined as Eu/Eu* (where Eu/Eu* ¼ Eu (N) /O(Sm (N) Â Gd (N) ), ranges between 6 and 228, with an average of 99 for the chimneys and 58 for the mounds (Fig. 3 and Table 2). The VDVF chimney and talus talc have 87 Sr/ 86 Sr ratios of between 0.706313 and 0.709168, respectively, which are similar to, but slightly less than, that of modern-day seawater (0.7092) 12 . This contrasts with the surrounding meta-gabbros that have 87 Sr/ 86 Sr ratios of between 0.702902 and 0.703657 (Tables 2 and 3).
The U-shaped REE patterns and positive Eu anomalies in the chimney and mound material indicate talc precipitation from VDVF vent fluid that has some similarities to high-temperature 'black smoker' vent fluid REE chemistry 13 . Reducing conditions in the VDVF vent fluid, indicated by the presence of sulphides, enhance Eu mobility by the formation of divalent Eu chloride complexes, especially during the dissolution of plagioclase, resulting in large, positive Eu anomalies in fluids and precipitates 14 . Furthermore, the VDVF talc 87 Sr/ 86 Sr ratios indicate precipitation following mixing of the high-temperature VDVF vent fluid with a high proportion of seawater (of at least 10:1). Chlorinity in the end-member VDVF fluid of 667 mmol kg À 1 is significantly elevated in comparison with seawater (546 mmol kg À 1 ), enhancing the complexation of LREEs in relation to mid-RREs (MREEs) and HREEs in hydrothermal solutions at high temperatures and pressures, resulting in elevated LREE talc /MREE talc ratios 15 . In contrast, moderate HREE enrichment is largely a crystallographic effect of the talc mineralogy, where HREEs substitute in the octahedral Mg site as a result of the ionic radii being of a more similar size compared with the LREEs 16 .
Other examples of seafloor talc deposits, reported from the St Paul and Conrad fracture zones, are inferred to have precipitated from the interaction of hydrothermal fluid with either seawater or a mafic protolith 17 . These talc samples have positive Eu anomalies and flat HREE profiles (Fig. 3), closely resembling those for the VDVF chimney and mound talus 17 . Seafloor talc deposits from elsewhere lack a positive Eu anomaly and have flat HREE profiles, consistent with formation as alteration products of an ultramafic protolith or sediment ( Fig. 3) 17,18 . In contrast, the positive europium anomaly for the VDVF talc is consistent with primary precipitation from hydrothermal fluids. The radiogenic 87 Sr/ 86 Sr ratios for the talc further indicate that a significant component of seawater is mixed with the vent fluid during talc precipitation.
We conclude from the petrographic and geochemical evidence that the VDVF talc is a primary precipitate from a hydrothermal fluid mixed with seawater. To date, no other seafloor, talc-dominated, active hydrothermal vent field has been reported, making the discovery of the VDVF a new and unique class of hydrothermal system.

Vent fluid chemistry.
To explore whether the VDVF is currently precipitating talc, we sampled and analysed the composition of the vent fluids escaping from three different chimneys with a maximum measured temperature range of 108-215°C (Table 3). It should be noted that the vent temperatures and fluid chemistry samples are decoupled, which precludes the possibility of extrapolating to an end-member temperature using fluid chemistry. Instead, we use the highest measured temperature of 226°C (ref. 19) and assume that this approaches the end-member temperature. We find that our vent fluid samples lie on a mixing line between seawater and zero Mg (Fig. 4). This is consistent with an end-member vent fluid from which Mg has been quantitatively removed in the subsurface 20 . When extrapolated to zero Mg, the VDVF end-member fluid has a moderate pH of 5.8 (at STP; Table 3) and a dissolved Si concentration of 7.5 mmol kg À 1 (ref. 21). End-member concentrations of K (17.5 mmol kg À 1 ) and Li (241 mmol kg À 1 ) are much higher than those generated by phase separation of seawater alone ( Fig. 4c and Table 3), indicating significant exchange with subsurface host rocks 20 . Strontium isotopes for the VDVF fluid samples also lie on a mixing line against Mg/Sr with seawater indicating an end-member vent fluid 87 Sr/ 86 Sr of 0.702908, close to the basement rock values of 0.702902-0.703657 ( Fig. 4d and Tables 2 and 3). Base-metal concentrations in the fluid samples range from 6.6 to 604 mmol kg À 1 for Fe, 4.6-14.2 mmol kg À 1 for Mn and 0.4-460.0 mmol kg À 1 for Cu, but do not show conservative mixing with seawater and hence end-members cannot be derived (Table 3). Variation in base-metal concentrations and ratios between hydrothermal vents across the VDVF suggests subsurface processes of precipitation and/or zone refining within the talc mounds. Owing to highly variable concentrations, reliable end-member vent fluid Fe and Mn concentrations could not be determined. However, the range of concentrations of Fe and Mn are 10-1,000 times lower than those reported for the Rainbow 'black smoker' end-member vent fluid 2 . In contrast, a chlorinity of 667 mmol kg À 1 for the end-member fluid is 22% higher than ambient seawater (546 mmol kg À 1 ), indicating a process of brine concentration.
The low metal concentrations as well as the near-neutral pH of the fluids venting at VDVF are consistent with temperatures in the water-rock reaction zone, which are significantly cooler than the B500°C calculated for 'black smoker' vents 22 . It is also known that elevated hydrogen sulphide concentrations in 'black smoker' hydrothermal vents are related to magmatic input 23 .
Hence, we suggest the low H 2 S concentrations at the VDVF are indicative of minimal magmatic contribution, consistent with the lower temperature of the end-member vent fluid and the ridge-flank setting of the vent field.
Increases in chlorinity for hydrothermal fluids interacting with an ultramafic basement have been suggested to occur in a number of ways: rock alteration, phase separation and brine and halite addition 2,24 . Temperatures in excess of 360°C are required to phase-separate seawater at 200 bar (that is, at the depth of the VDVF) and even higher for deep subsurface reactions 25 . This is considerably hotter than the maximum temperature recorded at the VDVF 21 and would require substantial subsurface cooling by seawater circulating deep within the talc mounds for which there is no evidence. The dissolution of residual halite, or the mixing with residual brine formed during an earlier and higher-temperature period of hydrothermal circulation and phase separation deep in the crust, could increase the end-member chlorinity 26,27 ; however, the VDVF and surrounding area lacks any mineralogical evidence for a 'black smoker'-like phase of venting. The process of serpentinization also has the potential to increase fluid chlorinity by the removal of water from the fluid 2,21 . Using the equation for the serpentinization of pure forsterite to serpentine and brucite (equation (1)), we calculate that the increase in chlorinity for the VDVF end-member fluid We have calculated a mass flux of end-member hydrothermal fluid at VDVF of B500 kg s À 1 (see below) for which the increase in chlorinity requires brine expulsion following serpentinization of pure forsterite at a rate of at least 470 kg s À 1 . Alternatively, the entrainment of an early-formed brine or halite phase is consistent with a cooling hydrothermal system, while serpentinization is compatible with the tectonic setting of the VDVF on slowly exhumed lower-crust and upper-mantle rocks. Both processes are ultimately constrained by the availability of residual brine, halite or fresh peridotite.
Fluid pH. A measured pH of 5.8 (at STP) for the VDVF end-member fluid is significantly higher than the observed range (pH 2.5-3.5) for sulphide-rich, high-temperature 'black smoker' vent fluids. It is also lower than the alkaline, low-temperature fluids vented at the ultramafic-hosted Lost City Vent Field 3,20 .    ARTICLE Hydrothermal fluids with intermediate pH (5)(6) can result from the interaction of seawater with mafic and/or ultramafic lithologies at temperatures o300°C (refs 28,29). Under these conditions, pH is controlled by the balance between Mg removal from seawater and silicate hydrolysis (equation (2)). At low water/rock ratios (o10), the rate of magnesium consumption is relatively low, and silicate hydrolysis acts as a pH buffer (equation (3)). At high water/rock ratios (450), H þ is produced at a rate greater than it is used up in hydrolysis reactions, resulting in lower pH 28 .
With its intermediate pH, 215°C temperature and setting on lower-crustal and upper-mantle rocks, the fluids venting at the VDVF are consistent with moderate-temperature interaction within a gabbro/peridotite basement (equation (3)).
Talc precipitation. Thermodynamic modelling using the Geochemist's Workbench and SUPCRT92 (refs 30-32), under the ambient pressure and temperature conditions of the VDVF, predicts instantaneous precipitation of talc and silica as the primary phases on mixing the 215°C VDVF end-member vent fluid with cold seawater (Fig. 5a). Both phases remain supersaturated throughout the mixing regime until seawater makes up B90%, and the fluid reaches B25°C (Fig. 5b). Below this temperature, talc remains undersaturated that, together with the kinetics of the reaction, may explain why talc was not visibly precipitating in the hydrothermal plume. Oscillatory zoning of talc and silica within active chimneys indicates a fluid composition at the VDVF that fluctuates around the intersection of the talc-silica saturation limits in the H 2 O-HCl-(Al 2 O 3 )-MgO-SiO 2 system. A fluid composition around the eutectic between talc and amorphous silica would be influenced by slight changes in Mg or Si activity, brought about either by precipitation of one or the other phase, or by fluctuating seawater proportions within the mound. Silica is saturated in the end-member fluid, therefore would precipitate when no seawater is available to provide Mg for talc precipitation. Such dynamic conditions, which are well documented at other sites [33][34][35] , could lead to alternating layers of mineral phases. Bands of chalcopyrite also indicate episodically higher temperatures. The prediction that talc precipitates directly as a result of mixing between the VDVF vent fluid and seawater is further supported by the range of 87 Sr/ 86 Sr ratios for the talc (0.706313-0.709168), which lie between the value of modern seawater (0.7092) and the VDVF end-member hydrothermal fluid (0.702908; Tables 2 and 3 and Fig. 4). The dissolved silica concentrations in the VDVF end-member fluids are similar to those reported from 'black smoker' vents 20 , where talc is also theoretically stable on mixing with cold seawater. However, the large quantity of sulphides precipitated at these vent sites results in talc and other silicates only occurring as accessory minerals (for example, Middle Valley) 36 . At the VDVF, the low metal content of the fluid results in only accessory amounts of metal sulphides, allowing talc and silica to become the dominant phases. By comparison, the higher pH of the Lost City vent fluids results in calcium carbonate and brucite precipitation instead of talc 3 .
In summary, the geochemistry of the solid and fluid phases and thermodynamic modelling are consistent with the observed texture of the talc, its growth into pore spaces, the presence of actively venting talc chimneys and the precipitation of significant volumes of talc-forming large conical mounds on the seafloor.
We suggest that similar conditions prevailed during the formation of the other, now hydrothermally extinct, talc mounds (for example, Mystic Mountain) located B700 m to the east of the currently active site. The current VDVF represents the latest stage in talc precipitation and hydrothermal circulation in the mafic and ultramafic basements beneath the Mt Dent OCC over tens of thousands of years.

Discussion
Hydrothermal venting on ocean crust more than 1 Ma in age has only been observed previously as either low-temperature (o10°C) diffuse flow of reduced fluids 37 or as mediumtemperature (91°C) venting of high pH fluids (for example, at the Lost City hydrothermal field 3,38 ) forming brucite and carbonate chimneys. In contrast, the VDVF vents fluid in excess of 200°C, from seafloor mounds and chimneys of predominantly talc, developed on a 1-2-Ma basement of lower-crust and upper-mantle rocks that were tectonically exhumed. This discovery, of an active hydrothermal system depositing predominantly talc as chimneys and forming large mounds on the seafloor, is unique. While talc-rich deposits of postulated hydrothermal origin have been reported before 11 , these were mainly precipitated as accessory phases associated with massive sulphides, either where Si-rich low-pH hydrothermal fluids mixed with seawater, or as a result of conductive heating of Mg-rich sedimentary pore waters. Only two seafloor talc deposits, from the St Paul's and Conrad fracture zones in the Atlantic Ocean 17 , appear directly comparable to the VDVF in terms of mineralogy, texture and geological setting.
At both sites, botryoidal talc, with similar REE patterns to the VDVF talc, was recovered from a tectonically exhumed gabbro-peridotite basement 17 . The presence of similar material to the active VDVF deposits at two different sites suggests that talc-dominated seafloor hydrothermal mineralization may be a widespread process at other slow spreading ridges, where tectonic uplift exposes lower-crustal and upper-mantle rocks to alteration by moderate-temperature hydrothermal circulation.
By measuring vertical velocities of vent fluid as it exits the seafloor, in situ fluid temperatures and the diameters of vent orifices at the VDVF, we have calculated a focused hydrothermal heat flux of 487 ± 101 MW, produced from the venting of B500 kg s À 1 of end-member fluid (Table 4). Despite being dissipated from crust of 1-2 Ma, this heat flux is comparable to that reported for focused heat flux dissipated by magma-driven 'black smoker' vent fields on zero-age ridge axes such as the Endeavour Ridge (302 MW), the East Pacific Rise at 21°N (200 MW) and the TAG vent field (758 MW) on the Mid-Atlantic Ridge 39 . When compared with other ridge-flank hydrothermal systems 37 , the heat flux at the VDVF is exceptionally high. It is also greater than the estimate for the Lost City hydrothermal field, which is thought to derive heat from the exothermic reaction of serpentinization 40,41 in addition to residual crustal heat 42 .
While hydrothermal activity is effective at extracting heat from the magmatic ridge axis 43,44 , processes of residual heat extraction from ridge flanks remain unclear. The intermediate temperature, near-neutral pH, low H 2 S and low concentration of dissolved metals in the VDVF vent fluid are all features consistent with hydrothermal circulation driven by residual heat within the tectonically uplifted lower crust and upper mantle 45 . Despite a high heat flux, the neutrally buoyant VDVF hydrothermal plume contains very little particulate matter 6 and, as a result, is difficult to detect using conventional optical sensors. Given the widespread occurrence of OCC's exhuming lower-crustal and upper-mantle rocks at ultraslow-medium spreading ridges, and the presence of similar talc deposits at the St Paul's and Conrad fracture zones 17 , we suggest that the processes leading to the formation of the VDVF could be widespread at slow-ultraslow spreading. In that case, the VDVF class of hydrothermal activity could contribute significantly to the cooling of the oceanic crust, but are not yet included in current estimates of global hydrothermal activity based on plume incidence 46 .

Methods
Sample collection. Sonar data, photographic and video imagery and samples of rocks, minerals, fluids and temperatures were acquired using remotely operated vehicles (Isis and HyBIS) during RRS James Cook cruises JC44 and JC82.
Vent fluids. High-temperature fluids were sampled using titanium gas-tight syringes at three different high-temperature chimneys (Table 3). Fluid exit temperatures were determined separately from the fluid sampling using a laboratory-calibrated, high-temperature thermal probe that was inserted deep into each vent orifice. Hence, the fluid samples and temperatures are decoupled. The highest measured value for each vent is reported in Table 3. Subsamples of vent fluids were taken for geochemical analyses, with pH, halides, sulphate and H 2 S being determined at STP (25°C) immediately after recovery of the titanium syringes on the surface. Anions were determined using ion chromatography, and cations using inductively coupled plasma atomic emission spectroscopy and inductively coupled plasma mass spectrometry (ICP-MS) at NOCS. Data quality was assessed using the 'Mottl Vent Fluid Database' protocols 47 . This included analysis of all solid precipitates from within the sampling syringes following total dissolution and weighted volumetric addition of concentrations to the clear fluid samples. Hence, any precipitates after the fluid samples were taken were redissolved and accounted for in final concentrations. Quality control included screening for samples with excessive Mg on a regression of elements such as Si, Ca, K and Li to zero Mg. Samples that were found to be contaminated with Mg concentrations more than those of seawater were subsequently excluded from the database. Our end-member concentration for dissolved silica is not as well constrained as that obtained in ref. 21, and thus their concentration is cited and used in our thermodynamic modelling.
Petrography. Petrographic analyses were conducted using transmitted and reflected light microscopy on polished thin sections. SEM was conducted using two different systems: one using a low-resolution Hitachi TM1000 Desktop SEM operating at 15 kV, imaging 1-cm 2 blocks of untreated broken surfaces; the second was a high-resolution LEO 1450-VP SEM operating at 20 kV and imaging carboncoated polished thin sections. Crystallographic analysis was made using X-ray diffraction with Cu Ka radiation 2y ranging between 2°and 76°, at steps of 0.02°a nd at rates of 0.02°per s. Unorientated air-dried mounts and glycolated mounts were scanned in the ranges 0-20°and 0-40°. Randomly orientated mounts were prepared and scanned in the range 57-62°to identify the peak signifying the (060) crystallographic plane.
Geochemistry. X-ray fluorescence (XRF) analysis was conducted on fused beads in a 10:1 ratio of sample powder to lithium tetraborate flux. Precision was determined from repeat analysis of standards UB-N, BRR-1 and OPY-1, and gives errors of o10%. Talc mineral separates were handpicked and washed in a weak acid solution and rinsed in de-ionized water. Following hydrofluoric and nitric acid digestion, the samples were analysed for trace element and REE analyses using ICP-MS. Standards included international reference materials: BIR-1, JB1a, JGB1, JB-3 and BHVO-2. Precision was determined from repeat analysis and gives errors in the range of 0.21-5.71%. Whole-rock analysis was conducted using both ICP-MS and inductively couple plasma atomic emission spectroscopy, and calibration was made using synthetic multielement standards. 87 Sr/ 86 Sr ratios were The equation used to calculate the heat flux is Q ¼ DT Â C Â M, where: Q, heat flux (Watts); DT, difference in temperature (°C) between the mean temperature of venting hydrothermal fluid and ambient bottom water (4°C); C, specific thermal capacity of seawater at 200 bar and 215°C (that is, 4,500 J kg À 1 K À 1 ) 53 ; M, mass flux (kg s À 1 ) ¼ A ÂV Â r, where A, area of vent orifice (m 2 ); V, mean vertical flow rate (m s À 1 ); r, density of seawater at 215°C. obtained using a thermal ionization mass spectrometer. Fluid and host rock samples were analysed by running through Sr-spec Tm resin to obtain 1 mg of strontium. Samples of strontium (200 mg) were obtained from pure talc by running through cation resin AG50-X8 200-400 and then through Sr-spec Tm resin. External reproducibility of analysis was checked by analysing standard NBS987 every 11 samples and is reported here with an average of 0.710250 (1s ¼ 0.0000053), which is within an error of the accepted value of 0.710254.
Thermodynamic modelling. Modelling was carried out using Geochemist's Workbench 30 at pressure and temperature conditions appropriate for the depth of the VDVF. A thermodynamic database for the composition of the vent fluid and seawater, at 250 bar and 4-215°C, with logK values calculated through this range of temperatures, was generated using DBCreate 31 , which in turn uses SUPCRT92 (ref. 32). Thermodynamic properties for aqueous species and phases were taken from the 2006 revision of the SUPCRT database 32 .
Heat flux calculation. Focused hydrothermal flow rates were calculated from high-definition video footage of the vertical velocity of particles entrained within venting hydrothermal fluid. Individual particles entrained in the exiting vent fluid were filmed over a rise height of B1 m and their velocity calculated from frame to frame. The distance travelled by the particles was measured against two parallel laser beams separated by a constant distance of 10 cm and shone through the rising vent fluid. The fluid exit velocities at each vent orifice were determined by measuring multiple particles in the upflowing fluid across the entire dimensions of the vent orifices. Different particles entrained in the up-flow zone from each orifice displayed only slight variation in their velocities, indicating that the flow rates are relatively uniform for fluids exiting each vent orifice (Supplementary Table 1). The velocity of a number of particles was thus calculated over a period of several minutes and a mean velocity derived for vent fluid as it escaped the vent orifice. The calculated error in the velocities is derived from the variation in particle velocity. These data were then integrated for each orifice as per established methods [48][49][50][51][52] .
This method models the vent orifices as circles, and makes the assumptions that the fluid is emanating from the vent at a constant velocity across its diameter, the temperature is uniform across the vent orifice and the particle velocities recorded were representative of the flow velocity. Specific heat capacity of the venting fluids was calculated using specific densities and thermal capacities for seawater 53 at the temperature and pressure conditions recorded at the VDVF.