Main

Methylmercury (MeHg) is a potent toxin that poses a threat to marine life and human health1, with expected global economic losses of US$19 trillion by 20502. The United Nations Minamata Convention on Mercury (Hg) aims to reduce human exposure to Hg by reducing anthropogenic emissions. Decades of research have comprehensively quantified recent global anthropogenic Hg emissions to 3,100 t per year (ref. 3). It is believed that anthropogenic Hg emissions have increased surface ocean Hg levels threefold since pre-industrial times4. Yet, to quantify how much anthropogenic Hg emissions have increased oceanic Hg levels, natural Hg inputs must be constrained. Atmospheric deposition, riverine inputs and submarine groundwater discharge are secondary Hg sources that recycle natural and anthropogenic Hg to the ocean. Hydrothermal vents release natural Hg into the ocean. Over 80% of the active volcanoes worldwide are oceanic5, yet their associated hydrothermal Hg fluxes remain poorly constrained. Indirect assessments via mass balance considerations proposed hydrothermal Hg fluxes between 360 and 3,720 t per year (refs. 6,7,8). The 2018 Global Mercury Assessment lowered the hydrothermal Hg flux from 600 t per year in the 2013 assessment to 100 t per year (ref. 3). Fluxes of Hg range from 20 to 2,000 t per year when scaling up direct hydrothermal vent fluid measurements from the four available studies at deep sites9. Total mercury (tHg) concentrations of the vent fluids used for this estimate range from 5 to 11,000 pmol l−1 (refs. 10,11,12,13).

Less is known about the fate of Hg in hydrothermal plumes when hydrothermal fluids mix with seawater, partially because early measurements of Hg in seawater near plumes were erroneous14,15. Two GEOTRACES transect cruises, crossing the slow-spreading Mid-Atlantic Ridge (MAR) and the fast-spreading East Pacific Rise, investigated Hg in hydrothermal plumes16,17. One sample above their seawater background levels (median 0.89 pmol l−1; Extended Data Table 1) with a Hg concentration of 13 pmol l−1 was reported at the MAR17. No change in Hg distributions was observed in the large plume of the East Pacific Rise16. Here we combine hydrothermal vent fluids, plume waters and rock samples from the Trans-Atlantic Geotraverse hydrothermal field (TAG) of the MAR to assess the fate of hydrothermal Hg inputs. We then use our findings to provide a refined estimate of hydrothermal Hg flux from mid-ocean ridges to the global ocean.

Mercury release from TAG hydrothermal vent site

We collected hydrothermal vent fluids, plume waters and seawater surrounding TAG on three dedicated oceanographic cruises. The unfiltered tHg concentrations of the vent fluid samples were 4,206 ± 262, 2,968 ± 87 and 367 ± 18 pmol l−1. During hydrothermal venting, magnesium (Mg) is quantitatively removed from seawater by precipitation as magnesium-hydroxy-sulfate hydrate. We used magnesium (Mg) concentrations of the vent fluid samples (4.59 ± 0.046, 32.9 ± 0.329 and 48.1 ± 0.481 mmol l−1) and their respective tHg concentrations to extrapolate the Mg-free vent fluid end member to 4,966 ± 497 pmol l−1 (Extended Data Fig. 1)18. The good correlation between tHg and Mg (R2 = 0.91) suggests that the tHg variations in the measured vent fluids are mainly due to mixing with surrounding seawater, with contributions from a single vent or fluid type. Previous studies at deep hydrothermal sites reported vent fluid tHg concentrations ranging from 5 to 11,000 pmol l−1 (refs. 10,11,12,13). Differences in geological settings may account for the variation in tHg concentrations of vent fluids. However, the vent fluid end member tHg concentration was not calculated in previous studies so different degrees of mixing with seawater may also explain the variability. The calculated vent fluid end member is around 6,000 times enriched in tHg compared with the median tHg concentration of the Atlantic Ocean (0.80 ± 0.42 pmol l−1; n = 1,917; Extended Data Fig. 2 and Extended Data Table 1).

Seawater and plume waters were collected during the HERMINE cruise (GEOTRACES GApr07) following the TAG plume flow dispersion in the water column. The tHg concentrations vary from 0.28 to 2.03 pmol l−1. The tHg concentrations are homogeneous from the surface to the deep waters at all stations (Fig. 1). In general, hot, acidic, low-density vent fluids are expelled through hydrothermal chimneys in buoyant plumes that rise in the water column mixing with seawater. Once the density of the buoyant plumes equals the density of seawater, the plumes reach neutral buoyancy and start moving laterally in non-buoyant plumes19. Non-buoyant plumes have been followed up to thousands of kilometres away from vent sites20,21. In this study, the definition of the magnitude and extent of the buoyant and non-buoyant plume was based on potential temperature, salinity, potential density and nephelometry (suspended matter) (grey shading in Fig. 1). The highest tHg concentration (2.03 pmol l−1) is observed within the buoyant plume layer at the TAG vent site (Fig. 1). High tHg concentrations (around 1 pmol l−1) are evident in the non-buoyant layer up to 10 km from the vent site (Fig. 1). The median of all the tHg values of this study including seawater and plume waters is 0.63 ± 0.077 pmol l−1 (n = 195) (Extended Data Table 1), a value slightly lower than the median of 0.80 pmol l−1 of the entire Atlantic Ocean, and in the range of 0.53–0.89 pmol l−1 reported for the North Atlantic Ocean where TAG is located (Extended Data Table 1)17,22. Concentrations of MeHg are low in the surface waters (0.05 pmol l−1 average above 500 m depth, n = 25), a peak around 1,000 m depth and consistently high concentrations at depth (0.28 pmol l−1 average below 1,000 m depth, n = 155) (Fig. 1). All plume waters and seawater samples display MeHg concentrations within previously reported range for the Atlantic Ocean (Extended Data Fig. 2 and Extended Data Table 1). Despite the high tHg concentration of the vent fluid end member, there is no detectable tHg or MeHg enrichment in the nearby waters (median tHg of 0.63 ± 0.077 pmol l−1 and median MeHg of 0.19 ± 0.12 pmol l−1). This may suggest that MeHg hydrothermal inputs at TAG are low, nor do the hydrothermal inputs stimulate MeHg production locally.

Fig. 1: Location of the TAG hydrothermal vent site sampled stations and vertical profiles for tHg and MeHg.
figure 1

a, Location of the TAG vent site (pink star), seawater literature data (black circles) and sampled stations (pink circles). Bathymetry from cruise M127 GEOMAR. b, tHg and MeHg vertical profiles. The grey shading shows the non-buoyant plume layer. Dotted lines and darker grey shading at the TAG station bound the depths of the buoyant plume. Maps in a created with QGIS (https://qgis.org/en/site).

Source data

Full size image

Dilution vs scavenging

Elements released from the hydrothermal vents can be diluted in seawater or scavenged onto sinking particles23. We used dissolved manganese (dMn) as an index to assess the importance of Hg removal processes following venting. Owing to its slow oxidation rate, dMn has been used as a conservative tracer of the dilution of vent fluids along hydrothermal plumes. During the cruise transit, particulate manganese (pMn) was lower than dMn with concentrations consistent with oceanic background levels24 (Extended Data Fig. 3). Average pMn considering all stations and depths within the plume was 2.5% of the total manganese (tMn) (Extended Data Fig. 4). Following previous studies19,25,26, we thus use dMn as a conservative tracer along the plume, being affected only by dilution. Combining manganese (Mn) data of the plume with our previously reported Mn concentration of the vent fluid end member at TAG (0.43 mmol l−1) (ref. 27), we calculate the dilution factor from the vent fluid end member to the non-buoyant plume (equation (1) and Extended Data Fig. 5). For comparison, dilution factors were calculated separately using tMn and dMn concentrations in the plume. No substantial difference was found between the two approaches, thus in further discussions, we refer to the dilution factor calculated with dMn (Extended Data Fig. 5). The average calculated dilution factor considering all depths with the plume and stations is 2.79 × 105 ± 1.4 × 103. The dilution factor is 2.36 × 104 at the TAG vent site and increases with distance to reach 1.21 × 106 at station 9 located 10 km away (Extended Data Fig. 5). Dilution factors are rather similar within the non-buoyant plume at each station.

To determine the relative importance of dilution and scavenging as the main process controlling the distribution of tHg, the estimated dilution factor was used to calculate expected tHg concentrations along our plume transect (Cn in equation (2)). Expected concentration (Cn) is the concentration that would be observed if only dilution drives changes in concentrations in the plume. Calculations were made using the Atlantic Ocean tHg median (0.80 pmol l−1; Extended Data Fig. 2 and Extended Data Table 1) and the tHg values reported in this study for the vent fluid end member and the plume waters (Extended Data Fig. 1 and Fig. 1). If measured tHg concentrations are equal to those expected (dilution efficiency = 100%; equation (3)), then tHg is behaving conservatively so that dilution is responsible for all the tHg losses from the vent to the plume. If observed tHg concentrations are below the expected ones (dilution efficiency < 100%), dilution cannot account for all the losses from the plume. Occasional dilution efficiencies above 100% may suggest that Hg is behaving more conservatively than dMn. Mean tHg dilution efficiencies per station range from 67 to 95% (Fig. 2). At the TAG station, the tHg mean dilution efficiency is 72%. The tHg dilution efficiencies tend to increase as the distance from the vent site increases, reaching 90% 10 km from the vent. Considering all plume depths and all stations, the dilution efficiency is 73 ± 20%, suggesting that tHg is largely conservative at TAG. Assuming that only dilution and scavenging account for losses from the plume, scavenging would remove the remaining 27% tHg.

Fig. 2: Dilution efficiency at different stations and depths.
figure 2

Blue shows tHg dilution efficiency. Black crosses indicate dFe dilution efficiency. The grey shading shows the non-buoyant plume depth. Dotted lines and darker shading at the TAG station bound the depths of the buoyant plume.

Source data

Full size image

To assess the validity of the dilution model, the dilution efficiency of dFe and tFe were also calculated. Calculations were made using the average seawater dFe (0.70 nmol l−1) (ref. 28), our previously reported Fe concentrations of the vent fluid end member (5.4 mmol l−1) (ref. 27) and plume Fe values19 (Extended Data Fig. 3). The average dilution efficiency considering all stations and depths within the plume is 14 ± 7% for dFe (Fig. 2) and 39 ± 8% for tFe (Extended Data Fig. 6). These low dFe and tFe dilution efficiencies are in good agreement with the previously reported non-conservative behaviour of Fe at TAG19,23.

Although Hg dilution efficiencies are relatively high, there is still potentially substantial deposition of Hg near the vent field. To further corroborate our observations, we analysed surface rocks and drill-core samples. The lithology of the active mound can be divided into six layers (Fig. 3a). The uppermost layer (Fe oxyhydroxides and cherts) displays Hg concentrations varying from 74 to 409 ng g−1 with a median of 112 ± 145 ng g−1 (circles in Fig. 3) and includes typical supergene lithologies including red ochres, Fe oxides, Mn-Fe oxyhydroxides and sulfides with different degrees of alteration (Extended Data Table 2). The massive sulfide layer displays concentrations from 99.9 to 1,187 ng g−1 when all the samples found in the layer are considered (Fig. 3 and Extended Data Table 3) but between 949 and 1,187 ng g−1 with a median of 1,021 ± 122 ng g−1 if just the typical massive sulfide samples (massive pyrite; Extended Data Table 3) are considered. The remaining layers, including anhydrite-rich, pyrite-silica, silicified wallrock breccias and chloritized basalt breccias, referred to as the stockwork zone, display Hg concentrations from 50 to 331 ng g−1 with a median of 90 ± 123 ng g−1. Hydrothermal sulfides (samples that lie in the bottom centre of the ternary diagram in Fig. 3b) show the highest Hg concentrations. The Hg concentrations are lower in silica-rich samples (upper part of the ternary diagram in Fig. 3b). This partitioning is due to the affinity of Hg for iron and sulfide phases rather than quartz, anhydrite or altered minerals in the host rock29. All samples display Hg enrichments in comparison to the typical oceanic crust concentrations30, with the surface and stockwork samples enriched up to ten times and the massive sulfide layer samples up to 100 times (Fig. 3c). The Hg concentrations of the typical massive sulfides recovered from TAG (purple and pink triangles in Fig. 3c) are within the range reported for other mid-ocean ridge areas (mean 1,991 ng g−1, n = 1,026)31.

Fig. 3: Hg concentrations of surface rock and drill-core samples.
figure 3

a, Lithology of the TAG mound deposit with the six layers and Hg concentrations of the drill-core samples. The depth of the drill-core samples is given in metres below the seafloor (mbsf). b, Ternary Fe-S-SiO2 diagram with all the rock samples. c, Hg concentrations of the rock samples from the surface, massive sulfide and stockwork layers. The numbers in the horizontal axis indicate the number of the sample as in Extended Data Tables 2 and 3. The black dotted lines indicate the Hg concentration of the oceanic crust30 and the median Hg concentration of hydrothermal sulfides at mid-ocean ridges31. Circles denote surface samples, triangles show samples from the massive sulfide layer and squares show samples from the stockwork zone. Note that the Hg concentrations are shown in a colour scheme.

Source data

Full size image

The best volume estimate of the TAG deposit is 2.95 × 106 t of rock at the active site, with 2.27 × 106 t in the massive sulfide layer and 6.8 × 105 t in the stockwork zone32. Using the median Hg concentrations measured for each layer (1,021 ng g−1 for the massive sulfide layer and 90 ng g−1 for the stockwork zone), the Hg in the deposit is estimated to be 2.38 t, with 2.32 t of Hg accumulated in the massive sulfide layer and 0.06 t Hg in the stockwork zone (part 1, Extended Data Table 4). Hydrothermal heat flux at TAG’s active mound was monitored during a seven-month interval, with a heat output varying from 55 to 86 MW (ref. 33). Older estimates present heat fluxes of up to 225 MW for the active portion of TAG34. Using the minimum and maximum heat flux estimates (55 and 225 MW), the volume flux of the discharged hydrothermal fluid at TAG is calculated to 0.04–0.14 m3 s−1 (equation (4), part 2, Extended Data Table 4)33,35,36. On the basis of the tHg concentration of the vent fluid end member (4,966 ± 497 pmol l−1) and the volume flux estimate, the discharge of tHg from the active TAG mound is estimated between 1,113 to 4,553 g per year (part 2, Extended Data Table 4). The activity at TAG has been proposed to occur on an intermittent basis, with 20,000 years of activity in the last 50,000 years of history37. The tHg release from TAG in all the active history of the site can be estimated between 22 and 91 t (part 3, Extended Data Table 4). If we consider that from the calculated 22 to 91 t of Hg that have been released from TAG in all the active history of the site, a minor fraction of 2.38 t (2.6–10%) has been retained in the deposit and that calculations indicate that dilution alone can explain the distribution of Hg from the vent to the plume (tHg dilution efficiency considering all the depths within the plume and all stations is around 73%; Fig. 2). We conclude with independent lines of evidence that Hg behaves conservatively at TAG. Thus, Hg released from TAG is able to persist long enough in the deep ocean and thereby contribute to the global Hg cycle (Fig. 4).

Fig. 4: Schematic model of Hg inputs at TAG.
figure 4

Seawater percolates the oceanic crust and is heated up reaching its maximum temperature close to the magma source. The vent fluid end member ascends through the hydrothermal vent where 10% of the Hg precipitates and stays in the rock. As the vent fluid is released from the vent, it starts mixing with seawater. Dilution with seawater accounts for most of the Hg losses from the plume with a maximum of 28% scavenging onto sinking particles at the TAG station.

Full size image

Implications for the global Hg cycle

We attempt to scale up our results by combining our findings with information from previous works to provide a measurements’ based estimate of hydrothermal Hg inputs. Submarine hydrothermal vents occur at tectonic plate boundaries including mid-ocean ridges, back-arc basins and submarine volcanic arcs. Hydrothermal sulfides from arc volcanoes display higher Hg concentrations than rocks from other tectonic settings (36,000 ng g−1, n = 241)31. Higher Hg concentrations at arc volcanoes may be attributed to sediment inputs during subduction and the subsequent release of magmatic volatiles into the hydrothermal systems. Back-arc settings and arc volcanoes can also occur at shallower depths than mid-ocean ridges. Shallow hydrothermal systems at arc volcanoes and back-arc regions generally release fluids of lower temperature and a higher volatile fraction, driving different Hg inputs9. Given these geological constraints, we use our results from TAG to provide an estimate for mid-ocean ridges exclusively. We use the latest global high-temperature mid-ocean ridge water flux estimate of (0.17–2.93) × 1013 kg per year (ref. 38) and the vent fluid tHg concentration of TAG (4,966 ± 497 pmol l−1) to calculate a mid-ocean ridge hydrothermal Hg flux between 1.5 and 32.1 t per year (part 4, Extended Data Table 4). Given that the highest tHg concentration of vent fluids reported at mid-ocean ridges is 11,000 pmol l−1 (ref. 10), the upper limit of the Hg hydrothermal flux estimate can be expanded to 64.7 t per year.

Recent models estimate that the annual Hg inputs to the oceans are 4,300–7,800 t per year from atmospheric deposition39 and 893–1,224 t per year from riverine discharge40, making a total of 5,193–9,024 t per year. The terrestrial natural volcanic and crustal degassing Hg inputs into the atmosphere have recently been revised to 335 ± 100 t per year (ref. 41). These natural terrestrial emissions are partly deposited in the ocean and partly in soils. Soil Hg can be sequestered for decades to centuries before running off via rivers to the oceans. Therefore, the anthropogenic inputs to the ocean can be calculated by subtracting the natural terrestrial flux (335 ± 100 t per year) from the total Hg input (5,193–9,024 t per year), resulting in 4,755–8,789 t per year (part 5, Extended Data Table 4). Given that mid-ocean ridges host the majority of known and projected hydrothermal vents, it is likely that they account for the majority of the global hydrothermal Hg flux42. Therefore, the flux of natural Hg to the ocean can be calculated as the sum of the terrestrial natural flux (335 ± 100 t per year) and the estimated mid-ocean ridge hydrothermal flux (1.5–65 t per year) to 237–500 t per year (part 5, Extended Data Table 4). The natural Hg flux to the ocean is therefore equivalent to 3–11% of the anthropogenic Hg flux to the ocean. The Minamata Convention on Mercury aims to reduce human exposure to Hg by reducing anthropogenic Hg emissions. As modern oceanic Hg levels seem to be predominantly due to anthropogenic inputs, we anticipate that the implementation of strict anti-pollution policies will lead to an effective reduction of oceanic Hg levels. Future research should focus on hydrothermal Hg inputs from arc volcanoes, back-arc settings, eruptive volcanism and off-axis venting to further constrain the global hydrothermal Hg flux to the ocean.

Methods

Samples from TAG, located at 26° 80′ N, 44° 50′ W and 3,620 m depth, were collected during three different oceanographic cruises. Vent fluids were sampled in 2012 during the SNAPMORE cruise onboard the R/V Knorr. Seawater and hydrothermal plume fluids were collected in 2017 during the HERMINE cruise (labelled GEOTRACES GApr07) cruise onboard the French R/V Pourquoi Pas?. Rock and drill-core samples were recovered in 1994 during the Ocean Drilling Program leg 158.

Sample collection

High-temperature hydrothermal vent fluids were collected unfiltered into gas-tight titanium samplers using the Remotely Operated Vehicle Jason II. Three samples for tHg from the vent fluids were collected and subsampled into 40 ml glass vials. All glassware used was acid cleaned with 3% v/v double-distilled HCl, rinsed with ultrapure water (18 MΩ cm, MilliQ), dried under a flow hood and then muffled at 450 °C for 4 hours. Surface rock samples (Extended Data Table 2) were collected by a series of dives of the submersible Alvin to the TAG active mound. Drill-core samples (Extended Data Table 4) are from drilling sites TAG1, TAG2, TAG4 and TAG5. As all the drilling sites are located in the active 200 m diameter TAG black smoker complex, all drilling locations are considered one site. More than one drill core was done at each location to recover as much stratigraphic section as possible. Cores were described and subsampled onboard. Rock samples were ground and homogenized with an agate mortar before Hg analyses. Seawater and plume waters were sampled by means of a trace metal clean carousel consisting of a polyurethane powder-coated aluminium frame attached to a Kevlar line. A conductivity temperature depth sensor (CTD, Seabird 911plus) was mounted on the trace metal clean rosette, and suspended matter data were processed in real time. The direction of the plume was monitored by a Lowered Acoustic Doppler Current Profiler deployed on a CTD/rosette before each trace metal clean cast. The trace metal clean carousel was equipped with 24 Teflon-lined 12 l GOFLO bottles with external closing. In total, 195 samples for tHg were collected unfiltered into 40 ml glass vials (Brooks Rand Instruments). For MeHg, 192 samples were collected unfiltered into 155 ml polyethylene terephthalate bottles (VWR) and acidified to 0.4% v/v with ultrapure trace metal clean HCl. All samples were stored at 4 °C until analysis.

Measurements

The tHg concentrations were measured within six months of sampling by a single gold trap custom-made purge and trap system43 following the US Environmental Protection Agency Method 1631. The vent fluid samples were vortexed before 1 ml was subsampled into 60 ml glass vials. To ensure the release of all Hg in the particles, 2 ml inverse aqua regia (VHCl/VHNO3 = ½) was added to the subsample. The subsamples were capped with a glass marble and heated to 120 °C for 8 hours. Recovery was evaluated using the reference material NIST-3133 (National Institute of Standards and Technology) (recovery > 99%) and blanks were run in ultrapure water (18 MΩ cm, MilliQ). The titanium syringe, used for vent fluid sampling, was blank tested by adding a background seawater sample. One subsample was taken a few minutes after adding the seawater to the syringe and a second the day after. The blank was determined to be 4.4 ± 1.9 pmol l−1 equivalent to less than 1% of the lowest vent fluid concentration measured in this study. For the seawater samples, oxidation was done using a BrCl solution. Potassium bromide (Sigma-Aldrich) and potassium bromate (Sigma-Aldrich) were heated for 4 h at 250 °C to remove Hg traces before preparing the BrCl solution with double-distilled HCl. A standard addition protocol was used to blank test the BrCl before adding it to the samples. Known volumes of BrCl were added to the same seawater sample, and the Hg concentration was measured. Increasing concentrations of Hg with the increasing volume of added BrCl would imply contamination of the BrCl solution. No contamination was found. After oxidation of the vent fluids and seawater, tHg was measured by cold vapour atomic fluorescence. To calibrate the tHg measurements, a seven-point calibration curve was constructed (R2 > 0.9998) spiking the reference material NIST-3133 into previously purged Hg-free seawater. Standard bracketing with NIST was done every six samples as a quality control to correct for the instrumental drift. The recovery was 94 ± 7%. The detection limit of the instrument was 0.03 pmol l−1 three times the standard deviation of the blank. Certified reference materials ORMS-5 (National Research Council Canada) and ERM-CA400 (European reference materials) were run at the beginning and the end of each measuring session to confirm the precision of our analytical method. Certified concentrations for tHg are 131 ± 6.5 pmol l−1 for ORMS-5 and 83.8 ± 5.5 pmol l−1 for ERM-CA400. Both CRMs were always within the certified ranges with 131 ± 3.1 pmol l−1 (n = 44) for ORMS-5 and 83.9 ± 3.5 pmol l−1 (n = 46) for ERM-CA400. Additional information on the uncertainty of the measurements is given in Supplementary Information 2.

MeHg was measured via a gas chromatography sector field inductively coupled plasma mass spectrometer and quantified by isotopic dilution. Briefly, a subsample of approximately 145 ml was transferred into 160 ml narrow-mouth glass vials (VWR). The glass vials were acid cleaned and muffled at 450 °C for 4 hours before use. The pH of the samples was adjusted to pH 3.9 using a sodium acetate buffer solution (ULTREX II Ultrapure Reagent, J.T. Baker) and NH3 (ULTREX II Ultrapure Reagent, J.T. Baker). Isotopically enriched spikes i199Hg and MM201Hg (ICS) were added to the samples targeting an optimal excess ratio of 4.2 and 5.1, respectively. After 10 minutes of equilibration, a freshly made 5% v/v solution of sodium tetrapropylborate (Merseburger Spezialchemikalien) and 150 μl isooctane (Sigma-Aldrich) were added, and the bottles were sealed with a Teflon-lined crimp cap (Fisherbrand). Samples were shaken for 15 minutes on an orbital shaker (Edmund Buhler KS15). The organic phase containing the derivatized Hg was transferred to vials for injection on a gas chromatography column (GC, THERMO GC 1300 with GC220 transfer module) coupled to a sector field inductively coupled plasma mass spectrometer (SF-ICP-MS, Thermo Element XR) system. The detection limit was 0.001 pmol l−1. Reverse isotopic dilution with Brooks Rand MeHgCl standard traceable to NIST1641E was carried out on the MM201Hg to determine the accuracy of the isotopically labelled spike, as compared to the theoretical concentration of 96.6 ± 7.9%.

The Hg concentration of the rock samples was analysed by cold vapour atomic absorption spectrometry (AAS, LECO AMA 254) according to the US Environmental Protection Agency (EPA) method 7473. Briefly, between 20 to 100 mg of each homogenized sample was introduced into a pre-cleaned nickel vessel. When dried and thermally decomposed, the generated products including Hg were carried away by oxygen flow to the catalytic decomposition tube and purified by the adsorbent. Then the Hg was selectively trapped by gold amalgamation, heated, desorbed and detected by AAS with a low-level optical cell. The instrumental calibration was made with the NIST-3133 certified standard solution and verified daily against the marine sediment certified reference material MESS-4 (certified concentration of 90 ± 40 ng g−1, National Research Council of Canada). The absolute detection limit of Hg was 0.01 ng. Quality assurance and quality control were conducted using method blanks, sample replication and certified reference material (MESS-4). Method blanks were <1% of the Hg mass of the measured samples. The MESS-4 was analysed after every five samples to test for recovery. The reference material was within the certified ranges with 70 ± 2.4 ng g−1 (n = 8). To confirm the complete release of Hg from the rock samples, the ash product of the thermal decomposition of two samples (Hg concentrations of 734 and 839 ng g−1) was further digested by adding inverse aqua regia and heating at 120 °C for 8 hours. The digested product was measured by cold vapour atomic fluorescence as described above. The Hg concentrations of the ashes after digestion were less than 14 ng g−1, indicating that more than 98% of the Hg is successfully released by the thermal decomposition method.

Calculations

The dilution factor DFn from the vent site to the buoyant plume was calculated for each depth n using equation (1) as previously defined by many authors35,44.

$${\mathrm{DF}}_{n}=\frac{{\mathrm{Mn}}_{\mathrm{vf}}-{\mathrm{Mn}}_{o}}{{\mathrm{Mn}}_{s}-{\mathrm{Mn}}_{o}}$$
(1)

Where the subscripts vf, o, and s indicate vent fluid end member, background North Atlantic Ocean seawater concentration, and sample concentration, respectively. The expected concentration of the tracer Cn (tHg, tFe or dFe) at each depth n was calculated using equation (2)35.

$${C}_{n}={C}_{o}+\frac{1}{{\mathrm{DF}}_{n}}({C}_{\mathrm{vf}}-{C}_{o})$$
(2)

This expected concentration (Cn) would be the concentration that would be measured if just dilution would account for the losses from the vent site to the plume. Therefore, dilution efficiency is defined as the percentage of the measured tracer (Cs = tHg, tFe or dFe) relative to the expected concentration (Cn) as

$$\mathrm{Dilution}\,\mathrm{efficiency}=\frac{{C}_{s}\times100}{{C}_{n}}$$
(3)

The volume flux (V) of discharged hydrothermal fluid was calculated using equation (4)33,35,36.

$$H=\rho\times{C}_{p}\times{V}\times\Delta T$$
(4)

Where H is the heat flux reported in the literature, ⍴ the density of the vent fluid 1,043 kg m−3, Cp the specific heat of hydrothermal fluid at constant pressure 4,100 J kg−1 K−1 and ΔT the temperature difference between the seawater 2 °C and the hydrothermal fluid 365 °C.